CADTH Reimbursement Review

Risdiplam (Evrysdi)

Sponsor: Hoffmann-La Roche Ltd

Therapeutic area: Spinal muscular atrophy (SMA)

This multi-part report includes:

Clinical Review

Pharmacoeconomic Review

Clinical Review

Executive Summary

An overview of the submission details for the drug under review is provided in Table 1.

Introduction

Spinal muscular atrophy (SMA) is a rare, autosomal recessive, neuromuscular disease and is the leading genetic cause of infant death.^1,2 The root cause in SMA is a deficiency in the survival motor neuron (SMN) protein, which is essential for the survival of motor neurons.^3,4 Specifically, the deficiency in the SMN protein leads to the degeneration of alpha motor neurons in the anterior horn of the spinal cord, causing irreversible loss of motor neurons and motor nerves, and progressive muscle weakness.¹ SMN protein is expressed mainly through the survival of motor neuron 1 gene (SMN1).^3-5 A second set of genes, the survival of motor neuron 2 gene (SMN2), can express the SMN protein, albeit at a drastically less effective rate than the SMN1 gene because only 10% to 15% of the protein produced by SMN2 is functional.^5,6

Traditionally, SMA is classified into 4 clinical subtypes based on the age of disease onset and the highest motor milestone achieved without disease-modifying treatment. These phenotypes differ in their presentations, severity of the disease, and prognosis. In SMA type I, patients show symptoms within their first 6 months of life, never achieve the motor milestone of sitting unsupported without disease-modifying treatment, and have a small chance of survival beyond 2 years of age due to respiratory failure without active symptomatic treatment and airway and respiratory management.^1,2,7-9 In SMA type II, patients achieve the milestone of sitting unsupported, but never walk independently without disease-modifying treatment. Symptoms generally appear between 6 months to 18 months after birth and most patients will survive past the age of 25,^7,10 with life expectancy improved by symptomatic treatment.¹⁰ SMA type III makes up about 10% to 20% of SMA cases¹¹ and manifests after 18 months of age. These patients are able to walk independently.¹⁰ Type IV makes up a very small proportion of SMA cases with symptom onset as adults, the mildest form of the disease.

At the time this report was drafted, there were 2 SMA disease-modifying therapies approved in Canada: nusinersen (Spinraza) and onasemnogene abeparvovec (Zolgensma). In Canada, nusinersen is indicated for the treatment of 5q SMA, while onasemnogene abeparvovec is indicated for the treatment of pediatric patients with 5q SMA with bi-allelic mutations in the SMN1 gene and either 3 or fewer copies of the SMN2 gene, or infantile-onset SMA.¹²

In addition to treatment with nusinersen, current standards of practice involve clinical monitoring and surveillance, anticipatory management of symptoms, and attempts to improve overall quality of life. Additionally, newborn screening for SMA is available in Ontario and may become available in other provinces and territories in the future, which would allow earlier identification and treatment of SMA, and earlier treatment improves the likelihood for patients to achieve treatment goals.

Risdiplam was submitted to CADTH with a proposed (pre-Notice of Compliance [NOC]) indication for the treatment of SMA. Risdiplam received an NOC from Health Canada on April 14, 2021, with a final indication for the treatment of SMA in patients 2 months and older. The product monograph notes that that are no data available in infants younger than 2 months of age and, therefore, risdiplam is not indicated in this patient population. It also notes that there are limited data on risdiplam for patients older than 25 years of age.

Risdiplam is a SMN2 pre-mRNA splicing modifier. Risdiplam corrects the splicing of SMN2 to shift the balance from exon 7 exclusion to exon 7 inclusion into the mRNA transcript, leading to an increased production in functional and stable SMN protein. Thus, risdiplam treats SMA by increasing and sustaining functional SMN protein levels.¹³ The dose of risdiplam is determined by age and weight as follows:

• 2 months to less than 2 years of age: 0.20 mg/kg

• 2 years or older and less than 20 kg of body weight: 0.25 mg/kg

• 2 years or older and 20 kg or more of body weight: 5 mg.

Stakeholder Perspectives

The information in this section is a summary of input provided by the patient groups and clinician groups that responded to CADTH’s call for input, from clinical experts consulted by CADTH, and the CADTH participating drug programs for the purpose of this review.

Patient Input

There are 2 patient input submissions for this review from Muscular Dystrophy Canada (MDC) and Cure SMA Canada (CSMAC). A disclosure of any conflicts of interest for the organizations is available on the CADTH website; both organizations are registered charities.

The submission was based on semi-structured interviews, virtual interviews, a focus group of 5 adult patients and 8 parent caregivers, and a survey of patients and caregivers that gathered 96 responses. All respondents lived in Canada and all data were collected between October and December 2020.

Six main themes were apparent, which have been listed in order of frequency reported: (1) an enormous impact on activities of daily living; (2) detrimental effects on breathing, swallowing, and mobility in particular; (3) significant dependence on caregiving supports; (4) loss of independence and control; (5) pain, age-related fatigue, and mental health; and (6) fear of falling.

Some of the major health concerns expressed by both patient groups included respiratory function (and illnesses like pneumonia), muscle strength, fine motor skills, falls and safety, nutrition (inability or losing ability to chew and swallow), voice and speech, mental and emotional health, and being easily fatigued. Transportation time and distance along with accessibility when out in public were noted as important considerations in day-to-day life. The desire to maintain or regain independence for as long as possible was common among responses, though there was still the constant fear of progressive loss of function and declining health. Living with SMA requires a high degree of dependence on both caregivers and equipment, additional therapy, and medical appointments, all of which lead to exhaustion for both patients and caregivers as well as increased strain on mental health and relationships.

Responders indicated challenges with treatment with nusinersen that included the intrathecal administration, the need to travel, the possibility of hospitalization, and the side effects experienced after a lumbar puncture. Being aware of risdiplam, respondents felt that a daily, oral treatment would have a positive impact on their lives if it meant fewer hospital visits, less strain on hospital resources and staff, a convenient and easily accessible treatment, and that patients and families would be allowed to have stable careers, education, and family lives.

Clinician Input

Input From Clinical Experts Consulted by CADTH

A panel of 7 clinical experts from across Canada was convened to characterize unmet therapeutic needs, assist in identifying and communicating situations where there are gaps in the evidence that could be addressed through the collection of additional data, promote the early identification of potential implementation challenges, gain further insight into the clinical management of patients living with SMA, and explore the potential place in therapy of risdiplam. The clinical experts stated that there is an unmet need for a therapy with evidence of effectiveness in the treatment of adolescent and adult patients with SMA. Currently, there is no evidence from a controlled trial of the efficacy of nusinersen in adult and adolescent patients, and several jurisdictions do not reimburse nusinersen for adult patients. In pediatric patients with SMA, there is an unmet need for an effective and non-invasive drug in patients who are not eligible for, do not have access to, or opt not to receive onasemnogene abeparvovec. As such, the expert panellists consulted expected that risdiplam will be the first-line treatment for patients who do not receive onasemnogene abeparvovec.

The clinical experts do not see a role for combination therapy. However, the clinical experts can see a place for risdiplam as a bridge between patient diagnosis and the administration of onasemnogene abeparvovec if patients are eligible for onasemnogene abeparvovec and the approval and administration of onasemnogene abeparvovec treatment would take longer to fulfill than risdiplam. Considering the systematic nature of risdiplam, the clinical experts expect to observe better improvements in bulbar function than with nusinersen, which may be a rationale for switching in the cases of some patients. However, considering that failure to respond is likely to be due to a state related to the disease activity and stage, rather than failure of therapy, assuming treatment is delivered reliably, the clinical experts do not necessarily think treatment with risdiplam would produce a different response in patients who have already deteriorated while on nusinersen. The role of subsequent treatment with risdiplam after treatment with onasemnogene abeparvovec is not clear; there is yet to be reliable evidence of the long-term duration of the effect of gene therapy with onasemnogene abeparvovec.

The clinical experts emphasized that, regardless of treatment strategy, the time since diagnosis or symptom onset and the age of the patient are some of the most important factors that would determine the extent of benefit that may be observed in patients with SMA. To this extent, the earlier any disease-modifying therapy is initiated, the more likely that the patient will maintain existing motor function or achieve new motor milestones. Patients who are most likely to benefit from risdiplam are those who might be too old for onasemnogene abeparvovec and have anatomic deformities that make nusinersen administration difficult, if not impossible. Risdiplam would essentially be the only treatment available for them. The clinical experts identified that patients with advanced stages of the disease, who may require invasive ventilation or who have a chronic gastrointestinal condition that might hinder drug absorption could be the least suitable for treatment with risdiplam.

Clinician Group Input

Clinician input was provided by the Neuromuscular Disease Network for Canada (NMD4C). The group is a pan-Canadian network launched in 2020 that aims to bring together clinical, scientific, technical, and patient expertise in neuromuscular disease.

The clinician group was largely in line with the clinical experts’ panel consulted by CADTH.

Drug Program Input

Input from drug programs explored the questions of concomitant use, factors influencing the clinical decision as to the choice of SMA therapy, assessment of ineffective treatment, and the potential use of a higher dose of risdiplam. The responses to the questions are in Table 4.

Clinical Evidence

Pivotal Studies and Protocol-Selected Studies

Description of Studies

One single-arm uncontrolled trial and 1 randomized controlled trial were included in this CADTH reimbursement review.

FIREFISH Part 2 (N = 41) is an ongoing, open-label, single-arm, phase III trial that investigated the efficacy and safety of risdiplam after 12 months of treatment in infants with SMA type I, with 2 copies of the SMN2 gene, and with no invasive ventilation. A total of 41 infants received risdiplam at an age-determined dose. These infants had an average age of 5.2 (standard deviation [SD] = 1.47) months and had 2 SMN2 gene copies, and their onset of symptoms was reported at a mean age of 1.64 (SD = 0.70) months. Average disease duration in FIREFISH Part 2 was reported at 3.59 (SD = 1.35) months. At baseline, |||| of the infants were able to control their head upright, while |||| did not demonstrate any motor milestone achievement, and 70.7% did not require any form of ventilatory support. The primary outcome was sitting for 5 seconds without support after 12 months of treatment, as assessed by the Bayley Scales of Infant and Toddler Development, Third Edition (BSID-III) tool. Other key secondary outcomes that were included in a statistical testing hierarchy were the proportion of patients who achieve a score of 40 or higher in the Children’s Hospital of Philadelphia Infant Test of Neuromuscular Disorders (CHOP INTEND) at month 12, the proportion of patients who achieve an increase of at least 4 points in their CHOP INTEND score from baseline at month 12, the proportion of motor milestone responders as assessed by the Hammersmith Infant Neurological Examination–Section 2 (HINE Section 2) at month 12, the proportion of patients alive without permanent ventilation at month 12, and the proportion of patients sitting without support for 30 seconds (item 26 of BSID-III) at month 24. No minimal important difference (MID) was identified for the BSID-III total score or for the CHOP INTEND total score, while HINE Section 2 had an estimated MID of more than 1 point.

SUNFISH Part 2 (N = 180) is an ongoing, double-blind, placebo-controlled, phase III trial that investigated the efficacy and safety of risdiplam after 12 months of treatments in patients with SMA type II or non-ambulatory patients with SMA type III and who are between 2 years and 25 years of age (inclusive). In SUNFISH Part 2, 180 patients were randomized on a 2:1 ratio to risdiplam and placebo, respectively. The mean age of enrolled patients was 9.9 (SD = 5.8) years in the risdiplam group and 10.3 (SD = 6.1) years in the placebo group. Patients belonging to the age group of 18 years to 25 years formed the smallest age group (11.7% in risdiplam and 13.3% in placebo), followed by the age group of 12 years to 17 years (25.0% in risdiplam and 26.7% in placebo). Most patients had 3 SMN2 gene copies (89.2% in risdiplam and 83.3% in placebo), while more than two-thirds were diagnosed as having SMA type II (70.0% in risdiplam and 73.3% in placebo). At baseline, |||| were able to stand in the risdiplam arm and |||| in the placebo arm. The primary outcome was change from baseline in the Motor Function Measure–32 items (MFM-32) score at month 12. Key secondary outcomes within the statistical testing hierarchy were the proportion of patients with a change from baseline MFM-32 total score of 3 or more at month 12, the proportion of patients with a change from baseline in the total score of the Revised Upper Limb Module (RULM) at month 12, the proportion of patients with a change from baseline in the total score of the Hammersmith Functional Motor Scale Expanded (HFMSE) at month 12, the proportion of patients with a change from baseline in forced vital capacity (FVC) at month 12, the proportion of patients with a change from baseline in caregiver-reported SMA Independence Scale (SMAIS) total score at month 12, and the proportion of individuals rated as “improved” on the Clinical Global Impression–Change (CGI-C) score at month 12. The sponsor proposed that 3 points or more difference on the MFM-32 score may indicate the acquisition of a new function or the improvement of several functions. The RULM score has an estimated MID of 2.9 points, the HFMSE score has an estimated MID of more than 2 points, and the SMAIS score has an MID of 1 to 5 points. No MIDs were identified for the rest of the outcomes.

In addition to the included pivotal study, this Drug Reimbursement Review included a summary of 3 phase II supportive studies: FIREFISH Part 1, SUNFISH Part 1, and JEWELFISH. FIREFISH Part 1 (N = 21) is an ongoing, open-label, single-arm, dose-finding, phase II trial that aimed to determine the appropriate therapeutic dose of risdiplam and reported on exploratory efficacy and safety findings at 12 months of treatment. SUNFISH Part 1 (N = 51) is an ongoing, 12-week, double-blind, dose-ranging, placebo-controlled trial that turns into a 24-month, open-label, single-arm trial. SUNFISH Part 1 aimed to assess the safety, tolerability, pharmacokinetics (PK), and pharmacodynamics (PD) of risdiplam in patients with SMA type II or type III aged from 2 years to 25 years; the results of exploratory efficacy outcomes and safety were reported at 12 months of the study’s duration. JEWELFISH (N = 174) is an ongoing, phase II, open-label, exploratory study to assess safety and tolerability of risdiplam in patients with SMA who received prior disease-modifying therapy. Only interim safety results are currently available from the JEWELFISH trial.

Efficacy Results

FIREFISH Part 2 achieved the primary outcome where 29.3% of infants (12 out of 41 infants) were able to sit without support after 12 months on treatment. This was contrasted with a natural history threshold of 5% (P < 0.0001). Of the reported secondary outcomes that are within the statistical hierarchy (at 12 months of treatment), 56.1% of infants (23 out of 41 infants) had a CHOP INTEND score of 40 or higher (P < 0.0001 against a performance criterion of 17%), 90.2% (37 out of 41 infants) have achieved an increase of at least 4 points in the CHOP INTEND score from baseline (P < 0.0001 against a performance criterion of 17%), and 78.0% (32 out of 41 infants) were considered motor milestone responders assessed through the HINE Section 2 (P < 0.0001 against a performance criterion of 12%). At month 12, 85.4% of patients were alive and did not require permanent ventilation (35 out of 41 infants). This outcome was contrasted with a predefined threshold of 42% (P < 0.0001). These efficacy outcomes are supported by the exploratory efficacy results at month 12 of treatment in FIREFISH Part 1. Efficacy outcomes beyond 12 months are not yet available.

SUNFISH Part 2 achieved its primary end point where patients who received risdiplam had a mean difference versus placebo of 1.55 points (95% confidence interval [CI], 0.30 to 2.81; P = 0.0156) in the change of the MFM-32 score from baseline. The first secondary outcome tested within the statistical testing hierarchy after the primary outcome was the MFM-32 responders (change of 3 points or more from baseline). This outcome showed that 38.3% of patients in the risdiplam arm (44 out of 115 patients) were considered responders, compared to 23.7% in the placebo group (14 out of 59 patients) with an odds ratio (OR) of 2.35 (95% CI, 1.01 to 5.44; P = 0.0469) for risdiplam versus placebo. Subsequently, the change in RULM score was tested, with a mean difference versus placebo of 1.59 points (95% CI, 0.55 to 2.62; P = 0.0028). Subsequently, 2 co-outcomes were tested: change from baseline in the total score of HFMSE, which failed to achieve statistical significance (mean difference = 0.58 points; 95% CI, –0.53 to 1.69; P = 0.3015) and change from baseline in best percentage predicted value FVC (mean difference = –2.05; 95% CI, –6.67 to 2.56; P = 0.3804). Patient- and clinician-reported outcomes, measured through the SMAIS and CGI-C tools, were next on the statistical testing hierarchy, but since the previous outcomes failed to achieve statistical significance, no additional statistical testing could be performed with control of the type I error rate.

The tool used for the primary outcome, MFM-32, did not have a well-established method of assessing the MID. However, the sponsor has indicated that a change of 3 points or more may translate into either the acquisition of a new function or the improvement in performance of several functions. Based on this, the primary outcome failed to demonstrate such a difference, both in point estimate and upper range of the 95% CI (MFM-32 change from baseline mean difference = 1.55 points; 95% CI, 0.30 to 2.81). Two other key motor function measurement tools used by the sponsor had published MIDs: RULM had an MID of 2.9 and HFMSE had an MID of 2. Results of the mean difference versus placebo in both of these outcomes also does not achieve the published MID.

Harms Results

In FIREFISH Part 2, at least 1 adverse event (AE) was reported in all enrolled infants. Upper respiratory tract infection was the most commonly reported AE (46.3%), followed by pneumonia (39.0%), pyrexia (39.0%), and constipation (19.5%). Serious AEs (SAEs) were reported in 58.5% of the infants (24 out of 41 infants); the majority of the SAEs were related to respiratory problems or respiratory infections. Three infants died during the study. Two were attributed to pneumonia and 1 to respiratory failure.

In SUNFISH Part 2, at least 1 AE was reported in 92.5% and 91.7% of enrolled patients in the risdiplam and placebo arms, respectively. Upper respiratory tract infection was the most commonly reported AE (31.7% in the risdiplam arm and 30.0% in the placebo arm), followed by nasopharyngitis (25.8% in the risdiplam arm and 25.0% in the placebo arm), pyrexia (20.8% in the risdiplam arm and 16.7% in the placebo arm), and headache (20.0% in the risdiplam arm and 16.7% in the placebo arm). SAEs were reported in 20.0% of patients who received risdiplam and 18.3% in patients who received placebo. Most of the SAEs were related to respiratory problems or respiratory infections.

Critical Appraisal

FIREFISH Part 2

In the absence of a concurrent control arm in the form of a placebo control or an active control, there is a potential for bias (i.e., an overestimate) in treatment effect estimated with risdiplam in the single-arm trial FIREFISH Part 2. Without a randomized comparison to a control group, natural fluctuations in the disease cannot be adjusted for, nor can the effects of known or unknown confounders. It should be noted that the sponsor method of determining a minimum threshold for testing the null hypothesis is not considered a valid control. While basing these thresholds on well-described literature may be useful to reflect the natural history of the disease, this approach has many limitations that arise due to variations in study characteristics, measurement methods, patients’ characteristics, standards of care, and other potentially unknown or unmeasured confounders. The sponsor attempted to determine the threshold by using the upper 90% CI in the selected natural history studies. However, no other method of adjusting for potential heterogeneities was attempted. These limitations involving the study design and the method used to establish the null hypothesis threshold led to a level of uncertainty regarding the extent of the magnitude of the observed clinical effect that is solely attributed to risdiplam.

Investigators, outcome assessors, and parents were aware that infants had received risdiplam. This may increase the risk of observer and respondent bias in assessing outcome measures. The potential risk of observer bias may be lower for objective outcomes, such as death and permanent ventilation. The sponsor video-recorded the assessment of motor milestone outcomes and used 2 independent central reviewers to determine the outcomes. However, the lack of blinding may be a concern for other outcomes, including health-related quality of life (HRQoL) outcomes.

FIREFISH Part 2 did not include infants who were pre-symptomatic, younger than 2 months of age, older than 6 months, or had more than 2 copies of SMN2. Other SMA infants who were not included in the study may have a more severe presentation of the disease (e.g., older SMA type I patients, patients requiring ventilatory support) and the results of FIREFISH Part 2 cannot inform on the extent of potential benefit offered by risdiplam for these patients.

SUNFISH Part 2

• The sponsor randomized patients in a 2:1 ratio to risdiplam or placebo, respectively. Potential challenges that may be associated with such an allocation ratio include the need for a larger sample size to capture differences in treatment effect, and the potential for reducing the effectiveness of blinding as investigators and assessors would be aware that the probability of being allocated to active treatment is twice that of control. Based on the primary end point of the study, it appears to be adequately powered given that statistically significant differences were observed. A reduction in statistical power due to the 2:1 randomization ratio could potentially have an effect on the secondary outcomes and subgroup analyses, considering that there were no a priori power calculations for the subgroups and secondary outcomes.

• The sponsor’s original plan aimed to randomize a total of 168 patients, stratified by age groups (2 years to 5 years, 6 years to 11 years, 12 years to 17 years, and 18 years to 25 years at randomization). However, only a maximum of 30 patients were to be included and randomized in the age group of 18 years to 25 years, while a minimum of 45 patients were to be randomized into each of the other age groups. The study ended up randomizing 180 patients, of whom only 22 were in the age group of 18 years to 25 years. This discrepancy in the number of patients in the age groups means that the age group of 18 years to 25 years did not contribute the same magnitude of effect to the assessed outcomes as the other age groups.

While adult SMA patients were included in the SUNFISH Part 2 confirmatory trial, they represented the smallest age group in the study (a total of 12.2%) and have, as such, contributed the least of these age groups to the overall efficacy results. The generalizability of the overall results, in turn, is lowest to the adult age group. This is further illustrated in the potential inconsistency of the magnitude of effect across age groups in all outcomes, as suggested by the point estimates of the treatment effect. In these age group assessments, the adult age group would show the lowest point estimates along with the widest CIs. Thus, the small representation of adult patients with SMA in the SUNFISH study, the high degree of uncertainty in the results of the subgroups analysis, and the pathophysiology of SMA indicate that the generalizability of the overall results of SUNFISH may be limited for adult patients with SMA.

Additionally, SUNFISH Part 2 excluded ambulatory patients. Considering the nature of SMA, where alpha motor neurons are irreversibly lost as disease progresses, patients with higher motor functions may have a greater number of alpha motor neurons than patients who have lost such motor functions. Ambulatory patients may thus exhibit a variation in the response compared to non-ambulatory patients, and the generalizability of the SUNFISH Part 2 results to this patient population may be limited.

Indirect Comparisons

Description of Studies

One sponsor-submitted indirect treatment comparison (ITC) was reviewed. The sponsor-submitted ITC compared risdiplam to nusinersen in 2 distinct patient populations: infantile-onset SMA (SMA type I) and later-onset SMA (SMA type II or SMA type III). An unanchored matching-adjusted indirect comparison (MAIC) was performed for the SMA type I population; the analysis included the FIREFISH study for risdiplam (pooled subgroup of Part 1 with Part 2) and the ENDEAR study for nusinersen. Study design between FIREFISH and ENDEAR was different as FIREFISH was a single-arm trial and ENDEAR was a double-blind, sham-controlled, randomized trial. However, inclusion and exclusion criteria were similar, and the individual patient-level data from the FIREFISH study were weighted to match the mean age at first dose, duration of symptoms, and mean CHOP INTEND baseline score of the ENDEAR study.

An anchored MAIC was used for later-onset SMA and included the SUNFISH study for risdiplam and the CHERISH study for nusinersen, with the placebo arm in the SUNFISH study and the sham arm in the CHERISH study acting as a common comparator. The model was adjusted for age at screening, baseline motor function score, and SMN2 copy number.

Efficacy Results

Within the SMA type I network, and subsequent to matching based on the mean age at first dose, mean disease duration at screening, and mean score on CHOP INTEND, an effective sample size of 36.5 was estimated out of the original 58 patients pooled into FIREFISH. The adjusted population had a higher proportion of female patients, a lower proportion of patients with ventilatory support, and a lower mean HINE Section 2 score than patients in the ENDEAR study. The results of the SMA type I unanchored MAIC suggest a hazard ratio of ventilation-free survival of risdiplam versus nusinersen of 0.20 (95% CI, 0.06 to 0.42), and an overall survival hazard ratio of 0.26 (95% CI, 0.03 to 0.66). Motor function assessment using the HINE Section 2 showed favourable results for risdiplam in the outcomes of motor milestone response, full head control, and sitting without support while the outcome of rolling was favourable in the direction of nusinersen. Two outcomes, sitting with and without support and standing, did not show a clear direction. However, all the HINE Section 2–related outcomes had wide CIs, indicating poor statistical robustness in the data.

Comparison in the SMA type II or SMA type III population showed, subsequent to matching, an estimated effective sample size of 28.3 in the risdiplam arm and 8.8 in the placebo arm. This small sample size translated into results with wide, and sometimes unrealistic, CIs in both the base-anchored base analysis and the network meta-analysis (NMA) sensitivity analysis.

Critical Appraisal

Unanchored MAIC depends on the strong assumption that all known and unknown effect modifiers and prognostic factors are accounted for within the model. This is a very strong assumption that was unlikely to be achieved. There were several factors that the sponsor-submitted ITC identified as important but did not adjust for in the MAIC due to lack of reporting. In addition, the differences in the study design between FIREFISH and ENDEAR are other factors that cannot be accounted for using MAIC. Even after weighting on the feasible factors, the weighted FIREFISH sample showed imbalances in several baseline characteristics compared to ENDEAR. As such, there is considerable uncertainty regarding the actual observed effect that is attributed to risdiplam. And since no attempt was made to assess residual bias, the magnitude of the bias due to effect modifiers and prognostic factors in the reported estimates cannot be estimated. This limitation is relevant to all outcomes that were estimated from the SMA type I evidence network.

The network that included patients with SMA type II or SMA type III was done through an anchored MAIC along with a sensitivity analysis using a Bayesian NMA approach. Due to large discrepancies in the inclusion and exclusion criteria, the sponsor-submitted ITC only used a subset of patients from the SUNFISH Part 2 study that would have been included in the CHERISH study, reducing the sample size of SUNFISH Part 2 by 62% and breaking randomization in SUNFISH Part 2. Subsequently, the weighting of the SUNFISH Part 2 sample produced an effective sample size in the risdiplam arm of 28.3. There were considerable differences in the eligibility criteria between CHERISH and SUNFISH Part 2. These differences included age range, restrictions on ventilatory support, restrictions on baseline motor function scores, and restriction over the existence of anatomic complications due to SMA. The sponsor-submitted ITC attempted to account for these differences through choosing a subset of patients for the SUNFISH Part 2 study that would match the inclusion and exclusion criteria of CHERISH. Only 38% of SUNFISH Part 2 was used and it showed several baseline imbalances between the risdiplam arm and the placebo arm. Upon weighting this subsample, differences in baseline characteristics persisted with the SUNFISH Part 2 subset and between the SUNFISH Part 2 subset and the CHERISH study. The resulting effective sample size was too small to provide a robust analysis, as reflected by the wide CIs.

Other Relevant Evidence

Description of Studies

FIREFISH Part 1 was a multi-centre, open-label, single-arm, phase II study of the safety, tolerability, PK, and PD of risdiplam for treatment of infants with SMA type I. The study enrolled 21 patients for a total 24-month treatment period followed by an additional extension period until risdiplam is commercially available or is no longer produced by the sponsor. Patients in FIREFISH Part 1 had a mean age of 5.81 (SD = 1.38) months at enrolment and 71.4% of patients were female. The majority (81%) were White and the rest were Asian. A proportion of patients was receiving ventilation support (14.3% received bilevel positive airway pressure [BiPAP] for fewer than 16 hours per day and 19.0% received ventilation prophylactically) and most patients (95.2%) were able to swallow.

SUNFISH Part 1 was a multi-centre, open-label, placebo-controlled, dose-ranging, phase II study of the safety, tolerability, PK, and PD of risdiplam in infant and young adult patients with SMA type II or SMA type III. The study enrolled 51 patients separated into 2 groups by age: 2 years to 11 years and 12 years to 25 years. During the initial minimum 12-week, double-blind, placebo-controlled phase, patients in the younger cohort could receive 1 of the 0.02 mg/kg, 0.05 mg/kg, or 0.25 mg/kg doses while those in the older cohort could receive either 3 mg or 5 mg doses. After 12 weeks, patients on placebo were switched to active treatment in the dose group they were assigned until pivotal doses were selected. The selected doses were 5 mg if the patient’s body weight was at least 20 kg and 0.25 mg/kg if body weight was less than 20 kg. All patients then received the pivotal dose for a 24-month treatment period followed by an additional extension period until risdiplam is commercially available or is no longer produced by the sponsor. In SUNFISH Part 1, patients aged 12 years to 25 years made up group A and patients aged 2 years to 11 years made up group B. The younger cohort had a mean (SD) age of |||||||| years while the older cohort had a mean (SD) age of |||||||||| years at screening. Just over half of all patients were female (||||) and |||| were White. Overall, |||| of patients were receiving fewer than 16 hours of BiPAP support per day (|||| and |||| in the younger and older cohorts, respectively) and |||| of the younger patients were not receiving pulmonary care while |||| of the older patients were not. Few patients (13.7%) were ambulatory: 19.4% of those aged 2 years to 11 years and 5% of those aged 12 years to 25 years.

Efficacy Results

For FIREFISH Part 1 at month 12, a larger percentage of patients in the higher risdiplam dose group achieved the motor function and development milestones than those in the lower-dose group. For instance, none of the of patients in the risdiplam lower-dose group and 41.2% in the higher-dose group were able to sit without support for 5 seconds (BSID-III, item 22). A lower percentage of patients showed improved motor skills as measured by the CHOP INTEND and HINE Section 2 in the 700 ng*hour/mL group compared to the 2,000 ng*hour/mL or less group. At the month-12 assessment, there was 1 death in each group and all remaining patients were alive without permanent ventilation. Few patients from either group required no respiratory support and most were able to feed orally. For patient- and caregiver-reported outcomes, measured by the ITQOL-SF47, there was a median change of 0 for both groups.

Some exploratory efficacy outcome data in SUNFISH Part 1 were available at month 12 and month 24. |||||||||||||||||||||||||||||||| |||||||| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||||| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||||||| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ||||||||||||||||||||

Harms Results

Nearly all patients (95.2%) in FIREFISH Part 1 had at least 1 AE. The most common were pyrexia (66.7%), upper respiratory tract infection (47.6%), and cough, diarrhea, and vomiting (28.6% each). SAEs occurred in more than half of patients (61.9%), with pneumonia being the most common (19.0%). Four deaths occurred: 1 in the 700 ng*hour/mL group and 3 in the 2,000 ng*hour/mL or less group. Safety outcomes for SUNFISH Part 1 similarly indicated that almost all patients (96.1%) had at least 1 AE. The 5 most common AEs were pyrexia (54.9%), cough (35.3%), vomiting (33.3%), upper respiratory tract infection (31.4%), and nasopharyngitis (23.5%). Overall, 29.4% of patients experienced at least 1 SAE, with the most common being pneumonia (5.9%). There were no deaths reported.

Critical Appraisal

FIREFISH Part 1 was an open-label, single-arm, dose-finding, phase II clinical trial that provided results on the safety and efficacy of 2 doses of risdiplam in 21 infants with SMA type I. Results from Part 1 were used to inform the dose used in Part 2 of the study. With no placebo or active control arm, it is possible that there was confounding of the treatment effect. Furthermore, without comparison to a randomized control group, natural fluctuations in the course of SMA cannot be adjusted for, nor the effects of known and unknown confounders. Considering the study’s main objective as a dose-finding study, efficacy outcomes are considered exploratory in nature.

SUNFISH Part 1 was a placebo-controlled, dose-finding, phase II clinical trial that provided information on the safety and efficacy of flat and weight-based doses of risdiplam in 51 patients with SMA type II and SMA type III. Results from Part 1 were used to inform the dose used in Part 2 of the study. Initially, the study was double-blinded for a minimum of 12 weeks, at which point patients receiving placebo were switched over to active treatment in their assigned cohort. The pivotal dose was selected and all patients were switched to receive that dose for 24 months. Unblinding occurred at the start of the open-label treatment period. With a limited placebo-control period, it is difficult to know the effects of confounders over time once all patients were receiving active treatment. The Motor Function Measure–20 items (MFM-20) was mistakenly administered to 7 patients who should have completed the MFM-32. As a result, the investigators removed these 7 patients, all of whom were in the 2-year-old to 11-year-old cohort, from efficacy analysis.

Generalizability of FIREFISH Part 1 and SUNFISH Part 1 to the general SMA population was not part of the studies’ objectives, considering the dose-finding nature of the studies and that outcomes were exploratory. In addition, the small number of patients and the single-arm study design make it difficult to generalize beyond the study populations. Results of these phase II studies should best serve as supportive evidence for the phase III confirmatory trials, FIREFISH Part 2 and SUNFISH Part 2.

Conclusions

By month 12 of treatment, patients in the FIREFISH Part 2 study were able to demonstrate the attainment of motor milestones, ventilation-free survival, and overall survival at levels that were statistically significant when compared to natural history thresholds. Patients in the SUNFISH Part 2 study did achieve the primary end point, at month 12 of treatment, with a statistically significant improvement in the MFM-32 score versus placebo, as well as in a secondary end point using the RULM score, though these differences may not have achieved an MID. Additionally, SUNFISH Part 2 failed to achieve statistical significance in the HFMSE and FVC outcomes, and no conclusive result can be determined for the outcomes of the SMA impact scale and the CGI-C “improved” rating. Respiratory and infection-related AEs were most prevalent across the studies and were a related cause of death for 3 patients who died in FIREFISH Part 2.

Indirect evidence suggested that risdiplam could potentially be favoured over nusinersen in ventilation-free survival and overall survival. However, the high level of uncertainty associated with serious limitations imposed by the method used in deriving the indirect estimates coupled with the lack of assessment for residual biases preclude concluding that risdiplam is more efficacious than nusinersen for these outcomes

Evidence gaps exist in the lack of efficacy and safety outcomes in pre-symptomatic SMA patients, SMA patients with prior experience with nusinersen or onasemnogene abeparvovec, infants with a disease duration longer than 6 months, and SMA patients who require permanent ventilation. Additionally, despite adult patients with SMA having been included in the SUNFISH study, there are several limitations associated with generalizing the overall result of SUNFISH to the adult SMA population.

Introduction

Disease Background

SMA is a rare, autosomal recessive, neuromuscular disease and was the leading genetic cause of infant death before the availability of disease-modifying treatments.^1,2 The root cause in SMA is a deficiency in the SMN protein, which is essential for the survival of motor neurons.^3,4 Specifically, the deficiency in the SMN protein leads to the degeneration of alpha motor neurons in the anterior horn of the spinal cord, causing irreversible loss of motor neurons and motor nerves, and progressive muscle weakness.¹ SMN protein is expressed mainly through the SMN1 gene.^3-5 A second set of genes, SMN2 gene, can express the SMN protein, albeit at a drastically less effective rate than the SMN1 gene, as only 10% to 15% of the protein produced by SMN2 is functional.^5,6 The most common form of SMA, 5q SMA, is the result of bi-allelic mutations in SMN1 on chromosome 5q13, where 95% of the cases are due to homozygous deletion and the rest are due to hemizygous deletion along with point mutation.^8,11 As the main source of SMN protein no longer functions in SMA, the number of SMN2 copies becomes an important disease modifier and is inversely related to the severity of SMA.^1,2,7,8

There is no Canadian national data on the incidence and prevalence of SMA in Canada. However, the most commonly reported estimate of the incidence of SMA is 10 in 100,000 live births and a prevalence of 5 in 100,000 individuals.^9,11 The sponsor of risdiplam estimates the incidence in Canada as 1 in 6,000 to 15,000 live births and the prevalence at 1 to 2 per 100,000.¹⁴

Traditionally, SMA is classified into 4 clinical subtypes based on the age of disease onset and the highest motor milestone achieved without disease-modifying treatment. These phenotypes differ in their presentations, severity of the disease, and prognosis. However, while the subtypes provide a convenient means of classifying patients, it should be noted that patients exist along a continuum of disease severity with overlap in symptoms between subtypes. In SMA type I, patients show symptoms within their first 6 months of life, never achieve the motor milestone of sitting unsupported without disease-modifying treatment, and have a small chance of survival beyond 2 years of age due to respiratory failure without appropriate treatment.^1,2,7-9 In SMA type II, patients achieve the milestone of sitting unsupported, but never walk independently without disease-modifying treatment. Symptoms generally appear between 6 months and 18 months after birth and most patients will survive past the age of 25,^7,10 with life expectancy improved by appropriate treatment.¹⁰ SMA type III makes up about 10% to 20% of SMA cases¹¹ and manifests after 18 months of age. These patients are able to walk independently.¹⁰ Type IV makes up a very small proportion of SMA cases with symptom onset as adults, the mildest form of the disease. However, while the subtypes provide a convenient means of classifying patients, it should be noted that patients exist along a continuum of disease severity with overlap in symptoms between subtypes. This spectrum is represented in Figure 1 as published in Talbot (2017).¹⁵

Figure 1: Continuous Spectrum of SMA Phenotype

MM = minimal manifestations; SMA = spinal muscular atrophy; SMN = survival motor neuron.

Source: Talbot K, Tizzano EF. The clinical landscape for SMA in a new therapeutic era. Gene Ther. 2017; 24:529 to 533. doi: 10.1038/gt.2017.52 Licensed under: https://creativecommons.org/licenses/by-nc-nd/4.0/

In SMA type I without disease-modifying treatment, patients show symptoms within their first 6 months of life, never achieve the motor milestone of sitting unsupported, and have a small chance of survival beyond 2 years of age due to respiratory failure.^1,2,7-9 SMA type I is the most common type of SMA, accounting for about 60% of SMA diagnoses.¹¹ Almost all patients with SMA type I have 2 or 3 copies of SMN2, giving rise to a broad range of phenotypes.¹⁶ Additional subtypes of IA (also called SMA type 0), IB, and IC have been proposed based on age of onset, with IA being the earliest and most severe subtype, presenting symptoms at birth with joint contractures, severe weakness and hypotonia, respiratory insufficiency, and a life expectancy of less than 6 months.^7,8 SMA type IB patients have symptom onset after 1 week of age but before 3 months of age, and SMA type IC have symptom onset between 3 months and 6 months of age. Muscle weakness in SMA type I is severe to the point where patients typically cannot perform antigravity limb movements and have no head control, though facial muscles are spared.¹ Fine motor skills are affected, with infants unable to grasp using their whole hand.⁴ Weakness in the intercostal muscles in combination with sparing of the diaphragm leads to paradoxical breathing and a bell-shaped chest.^1,8 Bulbar weakness results in difficulty swallowing and feeding, with risk of failure to thrive and aspiration.^1,8 Reflux and impaired cough and swallowing contribute to risk of aspiration and recurrent pulmonary infections.^1,2,7,8 A gastrostomy tube for feeding combined with nighttime and possibly daytime non-invasive ventilation with BiPAP can improve quality of life^7,8 and life expectancy.¹⁷ Aggressive intervention with a tracheostomy and permanent ventilation is also possible and can prolong life expectancy; however, this is a decision to be made by the family with the support of health care providers.^7,8 In 1 study that examined 1,966 patients in the Cure SMA database (with data available between 2010 and 2016), the median survival for type I SMA was 13.6 months.⁹ In a recent analysis of 307 treatment-naive patients, the median survival was 9 days in SMA type IA, 7.7 months in SMA type IB, and 17 years in SMA type IC.¹⁸

In treatment-naive SMA type II, patients achieve the milestone of sitting unsupported, but never walk independently. Some may lose the ability to sit unsupported over time. Symptoms generally appear between 6 months and 18 months after birth and most patients will survive past the age of 25 years,^7,10 with life expectancy improved by appropriate treatment.¹⁰ Type II patients represent about 20% to 30% of incident SMA cases and most SMA type II patients have 3 copies of SMN2.¹⁶ In addition to the inability to walk independently, common symptoms are fine tremors of the upper extremities, tongue fasciculation, joint contractures, and scoliosis.^1,8,10 Scoliosis and weak intercostal muscles can cause restrictive lung disease.⁸ There is a range in severity, with weaker patients requiring non-invasive ventilation.¹ Difficulty swallowing is less common than in type I patients and difficulty with feeding comes from masticatory muscle weakness.¹ In 1 study that examined 1,966 patients in the Cure SMA database (with data available between 2010 and 2016), the median survival for type II SMA was 59.9 years.⁹ In a recent analysis of 307 treatment-naive patients with SMA, patients with SMA type IIA had an end point–free survival probability at 40 years of 74.2% and at 60 years of 61.5%.¹⁸ The same analysis reported that patients with SMA type IIB had a relatively normal end point–free survival probability at 60 years of age.¹⁸

SMA type III makes up about 10% to 20% of SMA incident cases¹¹ and manifests between 18 months of age and adulthood. These patients are able to walk independently, though some may lose this ability over time, and typically have a normal life expectancy.¹⁰ Most type III patients have 3 or 4 copies of SMN2.¹⁶ An age of onset before 3 years (SMA type IIIA) is associated with estimated probabilities of 73%, 44%, and 34% of walking 10 years, 20 years, and 40 years after onset.¹⁹ In those with age of onset after 3 years (SMA type IIIB), the estimated probabilities are 97%, 89%, and 67% for walking 10 years, 20 years, and 40 years after onset.¹⁹ SMA type III patients have little or no respiratory weakness.⁸ Ambulatory patients may exhibit abnormal gait characteristics due to proximal weakness¹⁰ while patients who lose the ability to walk often develop scoliosis.¹

A very small proportion of SMA cases is type IV or adult-onset SMA, the mildest form of the disease. Although muscle weakness is present, these patients retain the ability to walk, have a normal life expectancy, and do not experience respiratory or nutritional issues.¹

The correlation between the genotype, specifically the number of SMN2 genes, and the clinical phenotype (SMA types) is probabilistic in nature. In 1 natural history study, of 39 patients diagnosed as SMA type IB or IC, 16 patients (41%) had 2 copies of the SMN2 gene, 21 patients (54%) had 3 copies of the SMN2 gene, and 1 patient (3%) had 4 copies of the SMN2 gene.¹⁶ The same study showed that out of 87 patients diagnosed as SMA type IIA or SMA type IIB, 2 patients (2%) had 2 copies of the SMN2 gene, 75 patients (86%) had 3 copies of the SMN2 gene, and 7 patients (8%) had 4 copies of the SMN2 gene.¹⁶ Of 66 patients who were diagnosed as SMA type IIIA or SMA type IIIB assessed in the same study, 1 patient (2%) had 2 copies of the SMN2 gene, 19 patients (29%) had 3 copies of the SMN2 gene, 40 patients (61%) had 4 copies of the SMN2 gene, and 2 patients (3%) had 5 copies of the SMN2 gene.¹⁶ The study also assessed 5 patients diagnosed as SMA type IV and of these patients, 4 (80%) had 4 copies of the SMN2 gene and 1 patient was not reported.¹⁶ A summary of these subtypes is presented in Figure 2, as published in Talbot (2017).¹⁵

Figure 2: Spinal Muscular Atrophy Clinical Classification

SMA = spinal muscular atrophy; SMN2 = survival of motor neuron 2.

Source: Talbot K, Tizzano EF. The clinical landscape for SMA in a new therapeutic era. Gene Ther. 2017; 24:529 to 533. doi: 10.1038/gt.2017.52 Licensed under: https://creativecommons.org/licenses/by-nc-nd/4.0/

Standards of Therapy

At the time of drafting this report, there are 2 SMA disease-modifying therapies approved in Canada: nusinersen (Spinraza) and onasemnogene abeparvovec (Zolgensma). In Canada, nusinersen is indicated for the treatment of 5q SMA, while onasemnogene abeparvovec is indicated for the treatment of pediatric patients with 5q SMA with bi-allelic mutations in the SMN1 gene and either 3 or fewer copies of SMN2 gene, or infantile-onset SMA.¹²

In addition to treatment with disease-modifying therapies, current standards of practice involve clinical monitoring and surveillance, anticipatory management of symptoms, and attempting to improve overall quality of life. SMA patients receive monitoring for growth, gastrointestinal function, and nutrition; respiratory complications; and orthopedic complications (i.e., scoliosis and/or contractures). These standards of practice include respiratory management for all children with type I SMA and some with type II SMA,⁸ secretion mobilization in patients with weak cough,^1,7 management of swallowing difficulties,¹⁰ and various multidisciplinary strategies for managing gross motor functions and spinal deformities (e.g., physiotherapeutic, orthopedic, surgical).⁷

With the introduction of nusinersen, an updated consensus statement on SMA standard care was published.²⁰ The updated consensus statement maintained strong emphasis on the multidisciplinary approach to addressing the various aspects of the disease and the involvement of various organs, and suggested a proactive approach to respiratory care, various approaches to acute manifestations of SMA, and the inclusion of nusinersen with an emphasis on monitoring potentially disproportional improvements in function due to the intrathecal administration route of nusinersen.²¹

Considering the existence of disease-modifying therapies, newborn screening of SMA to detect pre-symptomatic patients has been included in the Ontario newborn screening for SMA in its screening program,¹⁴ while Alberta has an ongoing pilot program for SMA screening.²² For patients with SMA who are identified through a newborn screening program, published clinical guidance suggests immediate treatment in infants with 2 or 3 copies of SMN2.²³

Drug

Risdiplam was submitted to CADTH with a pre-NOC indication for the treatment of SMA. An NOC was received during the course of this review (April 14, 2021) with a final indication for the treatment of SMA in patients 2 months and older. The product monograph notes that there are no data available in infants younger than 2 months of age and therefore risdiplam is not indicated in this patient population. It also notes that there are limited data on risdiplam for patients older than 25 years of age.

Risdiplam is a SMN2 pre-mRNA splicing modifier. Risdiplam corrects the splicing of SMN2 to shift the balance from exon 7 exclusion to exon 7 inclusion into the mRNA transcript, leading to an increased production in functional and stable SMN protein. Thus, risdiplam treats SMA by increasing and sustaining functional SMN protein levels.¹³ The dose of risdiplam is determined by age and weight as follows:

• 2 months to less than 2 years of age: 0.20 mg/kg

• 2 years or older and less than 20 kg of body weight: 0.25 mg/kg

• 2 years or older and 20 kg or more of body weight: 5 mg

Risdiplam underwent Health Canada priority review. The sponsor requested reimbursement as per the indication.¹⁴

Risdiplam has gained market approval from the FDA and the European Medicines Agency (EMA). The FDA indication is for “the treatment of spinal muscular atrophy (SMA) in patients 2 months of age and older.”²⁴ The EMA indication is “for the treatment of spinal muscular atrophy.”²⁵

Stakeholder Perspectives

Patient Group Input

This section was prepared by CADTH staff based on the input provided by patient groups.

About the Patient Group(s) and Information Gathered

CADTH received 2 patient input submissions for this review from MDC and CSMAC. A disclosure of any conflicts of interest for the organizations is available on the CADTH website.

MDC is an organization that supports patients and caregivers affected by neuromuscular disorders, including muscular dystrophies and related muscular diseases. The organization represents more than 12,000 patients and nearly 19,000 family members and caregivers in Canada. MDC’s mission is to improve the lives of those impacted by neuromuscular disorders. CSMAC is a national registered charity that supports patients and families affected by SMA from diagnosis throughout life, and after loss of life. CSMAC’s mission is to improve both the quality and quantity of life for patients with SMA. Both organizations help fund research and offer programs and services for education, access to financial support for critical equipment, peer-to-peer networking, information on new treatments, medical advances, clinical trials, and advocacy.

MDC conducted interactive, semi-structured, virtual interviews with patients and caregivers to learn about their experiences with SMA. Between November 10 and December 4, 2020, 92 interviews were conducted.

CSMAC collected information for its submission through semi-structured interviews, a focus group, and a survey of patients and caregivers. All respondents lived in Canada and all data were collected between October 2020 and December 2020. The focus group consisted of 5 adult patients and 8 parent caregivers of patients with SMA. The survey was available in both English and French, and a link to it was distributed to the patient community in CSMAC’s database. Of the 96 responses, 17% represented patients with SMA type I, 58% represented patients with SMA type II, 24% represented patients with SMA type III, and 1% were other. Forty-seven responses were for adult patients, 20 were for patients from 11 years old to 20 years old, 16 were for patients from 5 years old to 10 years old, 8 were for patients from 2 years old to 5 years old, and 5 were for patients younger than 2 years of age.

Disease Experience

MDC asked patients to describe how SMA impacts day-to-day life and quality of life. Six main themes were apparent, which have been listed in order of frequency reported: 1) an enormous impact on activities of daily living; 2) breathing, swallowing, and mobility are mostly affected; 3) significant dependence on caregiving supports; 4) loss of independence and control; 5) pain, age-related fatigue, and mental health; and 6) fear of falling.

Some of the major health concerns expressed by both patient groups included respiratory function (and illnesses like pneumonia), muscle strength, fine motor skills, falls and safety, nutrition (inability or losing ability to chew and swallow), voice and speech, mental and emotional health, and being easily fatigued. Transportation time and distance along with accessibility when out in public were noted as important considerations in day-to-day life. The desire to maintain or regain independence for as long as possible was common among responses, though there was still the constant fear of progressive loss of function and declining health. Living with SMA requires a high degree of dependence on both caregivers and equipment, additional therapy, and medical appointments, all of which lead to exhaustion for both patients and caregivers as well as increased strain on mental health and relationships.

The following is a selection of quotes from patients and caregivers to illustrate living with SMA:

• “I have to have a wheelchair vehicle in order to get anywhere and then not only that, but I worry about how accessible the place is going to be once I get there – which is out of my control. What is accessible to one person is not access to the max. Especially if you are in a wheelchair full-time.”

• “Daily, I rely heavily on the assistance of PSW’s to meet all of my physical needs from getting dressed, toileting, transferring to my power chair, shifting several times to get comfortable, washing my face, fixing my hair, getting breakfast…the list goes on and on. Basically, I require help to do all things physical, even putting my arms around my son to give him a hug.”

• “I need 24/7 care so I can’t live on my own unless it’s in a facility which I never want. So I live with my parents who are pushing 60 and when I don’t have my personal care attendants, they struggle to take care of me.”

• “Mobility limitations takes a toll on mental health and happiness, it makes you feel shameful. I cannot enjoy my children’s hockey games and I feel excluded from social events and sporting events.”

• “With loss of function, he said at age 11 ‘This is no life, a life not worth living.’”

Experiences With Currently Available Treatments

In general, the major barriers respondents faced include the high cost of treatment and transportation as well as differing access or restrictions among jurisdictions. Caregivers also highlighted the stress that hospital visits caused patients, particularly children, which could lead to anxiety; the sacrifice made by patients and caregivers to take time off for travel, treatment, and recovery; the risks or side effects of treatments; and the lack of insurance coverage for medication. The method of administration was also a concern, and each has its own drawbacks. For example, spinal injection is invasive and would not be an option for patients with spinal fusion while inaccessibility to veins makes IV treatment a challenge.

Patients who had received Spinraza (nusinersen) every 4 months felt that it was a painful, risky, and invasive procedure. It was also noted that this treatment is not always successful and has led to cerebrospinal fluid leaks, headaches, and periods of regression in-between injections for some patients. Moreover, it may be an inappropriate choice since it relies on being able to find access points for administration. Although there is fear around the treatment process itself, patients and caregivers still worry about missing an injection and the impact it would have on the patient.

• “Accessing Spinraza initially was a rollercoaster. Without the recommendation for adults, Saskatchewan has determined approval on a case-by-case basis. It took a long time to get provincial approval…. But then not only that, maintaining access to treatment is almost just as difficult. Testing needs to be done with the Hammersmith test and upper extremity test every four months…. One-point difference from the previous testing and I could no longer qualify for Spinraza. That is a constant worry.”

Patients and caregivers shared some of the improvements they’ve seen with new medications:

• “My energy level has gone higher than it’s ever been. I have more trunk control and strength and I’m doing things that I haven’t been able to do in 10 years or more. Driving is easier, baking is easier, getting through the day is easier and I no longer need a nap halfway in between.” (adult who received nusinersen)

• “The day after she had her infusion she was able to sit unassisted for over 30 minutes which is something that she had never been able to do before. We saw gains almost right away and we have seen prolonged improvement as well as more energy and movement.” (parent of patients who received onasemnogene abeparvovec)

Despite the benefits that treatment can offer, many struggle with gaining access:

• “I am not currently on any specialized treatment for SMA. This is due to the fact that there are very few treatments for SMA and they are difficult to access. This is because of prohibitive costs, lengthy and restrictive approval processes, invasive or difficult route of administration, and travel to medical facilities.”

• “I am currently not receiving any treatments for SMA. This is due to the current costs of treatments being well outside my financial means and local government not funding treatment for people over 18 currently.”

Improved Outcomes

Many respondents stated that they wanted to see a halting of disease progression and are hopeful for improvements in respiratory ability, mobility, strength, and feeding. Overall, patients and families valued greater independence, longer life expectancy, and improved quality of life, particularly for the patient, but also within the family unit. For new treatments, respondents would like to see ones that are less invasive, have fewer side effects, and are covered through insurance.

• “It would be great if a treatment would stop progression and allow me to rebuild muscle and strength and give me everything I desire to do.”

• “Stopping the progression is critical, I can still drive a little bit, I love writing and would love to be able to gain a bit of ability to be able to do that. Maintain my eating, would love to be able to stay in my own home and die there, not have to go to LTC.”

Being aware of risdiplam, respondents felt that a daily, oral treatment would have a positive impact on their lives if it meant fewer hospital visits, less strain on hospital resources and staff, a convenient and easily accessible treatment, and that patients and families would be allowed to have stable careers, education, and family lives.

Experience With Drug Under Review

MDC and CSMAC collected responses from 4 patients and 6 patients, respectively, who had previous experience with risdiplam through clinical trials or special access programs. One patient interviewed by MDC had been receiving risdiplam for 2.5 years while the patients from the CSMAC group had been on risdiplam for between 4 months and 2.5 years and were from 3 years to 13 years old.

In general, patient experience was very positive with improvements in motor and respiratory function, strength, energy, appetite, and weight gain, along with decreased illness and dependence on medical equipment. Respondents were pleased that administration could be done at home and that the medication was available from a pharmacy rather than the hospital. When compared to nusinersen, multiple patients and families noted the benefit of being able to avoid the waning effect between doses, have fewer side effects of treatment, and reduce the potential risks of repeated spinal punctures. A few respondents stated that with risdiplam, it took several months to see a noticeable change and that some patients experienced acid reflux that decreased over time.

• “All of our experiences have been positive with this treatment. His health has improved in way that we didn’t think it would.”

• “…in July of 2020 prior to Risdiplam he was able to sit unassisted out of his brace for 10-15 seconds on occasion, and in his brace for between 30 seconds to 1 minute. Now, in November 2020, he can sit unassisted without his brace for 1-2 minutes on occasion, and in his brace he can sit unassisted for anywhere from 15-30 minutes routinely.”

• “It has given us hope for our child’s future. We have less worry because he is gaining muscle strength instead of losing it. He has been able to fight off illness, so we have less fear of him having to go to the hospital. Knowing that a medication has saved your child’s life positively changes everything. ”

Companion Diagnostic Test

Of the 92 responses from the MDC group, 84 stated having a genetic test to confirm SMA diagnosis via blood test. Respondents felt it was a simple procedure, did not cause significant anxiety, was at no additional cost, and provided quick results. While most felt they did not face significant barriers, a few stated that having other family members tested (e.g., siblings of a child with SMA or parents when family planning) was a major hurdle. For instance, 1 parent recounted their experience when attempting to have their other children tested: “I pushed for testing for them as well, and this proved to be extremely frustrating and time consuming for me…. I still do not understand why I encountered so much resistance trying to get these tests performed.”

Additional Information

A few family-centred considerations could be noted in the patient submissions: families having to find care for other children during appointments, the importance of being able to care for all their children equally, having treatments and a lifestyle that improved quality of life for the whole family, and caregivers having the time and energy for things that often came second to caring for their family member with SMA.

• “My other kids would not have to deal with being left with other people while I take her to treatments, and I would be way less stressed and better able to better distribute myself among them, avoiding the hypervigilance and focus on my affected child after each treatment that disrupts our family dynamic. Risdiplam would mean WAY less trauma and anxiety for my family and a better quality of life for my daughter, my spouse and I, and for my other 3 small children.”

Respondents also emphasized issues with restrictions on age and ventilator status.

• “Do not limit treatment because a child is ‘invasively ventilated.’ He has shown that he can make huge improvements with compassionate access to these drugs, but without compassionate access he would have been locked out of any provincially funded options because he is invasively ventilated.”

Both patient groups clearly expressed that with access to effective treatments, they were hopeful for patients’ futures.

• “Parents who have children with SMA are excited about the possibility of having more than one treatment option. Our kids are intelligent, and deserve a chance at a life with greater mobility, and easier breather, and speaking. Given the opportunity to make it to adulthood they will become major contributors to our society.”

• “No child or adult should have to wait. It’s a literal matter of life and death for them. The benefits to these SMA patients are enormous too and, if not life-saving, certainly life-changing.”

Lastly, CSMAC emphasized the importance of having new therapies for adult patients with SMA since many are not accessing treatment in Canada and highlighted that these medications can have life-changing results, allowing patients to live their best lives at any age.

Clinician Input

All CADTH review teams include at least 1 clinical specialist with expertise regarding the diagnosis and management of the condition for which the drug is indicated. Clinical experts are a critical part of the review team and are involved in all phases of the review process (e.g., providing guidance on the development of the review protocol, assisting in the critical appraisal of clinical evidence, interpreting the clinical relevance of the results, and providing guidance on the potential place in therapy). In addition, as part of the risdiplam review, a panel of 7 clinical experts from across Canada was convened to characterize unmet therapeutic needs, assist in identifying and communicating situations where there are gaps in the evidence that could be addressed through the collection of additional data, promote the early identification of potential implementation challenges, gain further insight into the clinical management of patients living with SMA, and explore the potential place in therapy of risdiplam (e.g., potential reimbursement conditions). A summary of this panel discussion is presented as follows.

Unmet Needs

In the pediatric population, nusinersen has limitations associated with the route of administration, the distribution of the active substance within the body, an underwhelming effect on respiratory and bulbar function, and a growing clinical experience with diminished response, or a reversal of effect over time. The clinical experts noted that the route of administration of nusinersen can be associated with several concerns, including uncertainty as to whether the appropriate dose has been administered (in part due to cerebrospinal fluid leaks that may contain nusinersen), unequal distribution of nusinersen in the central nervous system (CNS), and the small risk of nerve damage in infants and toddlers. Also, some older patients may have spinal deformities (e.g., scoliosis) that may prevent them from receiving nusinersen. In addition, patients with potential allergies or who have a bleeding disorder would not be eligible for nusinersen. The introduction of onasemnogene abeparvovec may alleviate some of the limitations associated with nusinersen. However, considering the importance of prompt treatment, there remains a group of patients that may need immediate access if onasemnogene abeparvovec is not readily available. In addition, there are patients who might have contraindications to onasemnogene abeparvovec due to potential hepatotoxicity or the existence of anti–adeno-associated virus 9 (AAV9) antibodies, or opt out of treatment with onasemnogene abeparvovec. There is also limited evidence of the benefit of onasemnogene abeparvovec in older children.

Overall, in the pediatric population, there still exists a need for reliable, effective, and non-invasive treatment.

Treatment of adult patients with SMA is limited to nusinersen. However, there is limited evidence of its efficacy and safety in adult patients with SMA. In addition, access to nusinersen across Canada’s provinces and territories for adult patients with SMA is variable. There remains an unmet need for an effective drug in the adult population that can clearly demonstrate benefits in this age group with slow disease progression.

Finally, there remains a need for therapies that may reverse the effect of SMA and therapies that can prove beneficial for patients with invasive ventilation. Currently, all available treatment options can only work on viable alpha motor neurons but cannot reverse the damage already inflicted.

Place in Therapy

Currently, all pediatric patients are treated with nusinersen. With onasemnogene abeparvovec availability, most patients who would be eligible for onasemnogene abeparvovec are likely to be treated with it, unless there exists a contraindication (e.g., liver disease), high titres of anti-AAV9 antibodies, or the refusal of the family to use gene therapy. In SMA patient populations that may not have access to onasemnogene abeparvovec, reimbursement eligibility would determine whether nusinersen or risdiplam would be administered. Should the patient have access to both, it is likely that the oral route of risdiplam administration would be preferable over the intrathecal route of nusinersen to eliminate potential AEs and limitations associated with the nusinersen intrathecal route and overcome some of the administration challenges in patients with scoliosis or contractures. The role of risdiplam in pre-symptomatic patients is not yet clear pending more evidence regarding the efficacy and safety of risdiplam in the pre-symptomatic population.

Pediatric patients with later-onset SMA, who might be too old for onasemnogene abeparvovec, are likely to start treatment with risdiplam. In essence, the clinical experts expect risdiplam to replace nusinersen due to potential differences in the risk-benefit ratio.

Risdiplam will likely be the first-line therapy for adult and adolescent patients, largely due to the paucity of evidence for nusinersen efficacy in this group of patients, the lack of reimbursement of nusinersen in some jurisdictions for adult patients, the systematic distribution of the therapy, the faster time to reach therapeutic concentration, as well as the non-invasive route of administration of risdiplam compared to the intrathecal route of nusinersen.

The clinical experts do not see a role for combination therapy in clinical management of SMA. There is no rigorous and/or longitudinal evidence supporting combined therapy. Moreover, the clinical experts see no biological rationale for combination therapy, irrespective of medication. This is considering that all agents act on upregulating the same protein (SMN), and as long as the delivery of the active ingredients to the target cells is proper, combined therapies would be expected to be redundant.

The clinical experts can see a role where risdiplam is given as a bridge until the patient can be treated with onasemnogene abeparvovec. This takes into consideration the importance of prompt treatment to avoid any irreversible loss of alpha motor neurons. Considering the systematic nature of risdiplam, the clinical experts expect to observe better improvements in bulbar function than with nusinersen. However, the clinical experts do not necessarily think treatment with risdiplam would produce a different response in patients who have already deteriorated while on nusinersen. The role of subsequent treatment with risdiplam after treatment with onasemnogene abeparvovec is not clear; there is yet to be reliable evidence measuring the long-term duration of the effect of gene therapy with onasemnogene abeparvovec.

Patient Population

The clinical experts emphasized that, regardless of treatment strategy, time since diagnosis or symptom onset is the most important factor that would determine the extent of benefit that may be observed in patients with SMA. To this extent, the earlier any disease-modifying treatment is initiated, the more likely that the patient will maintain existing motor function or grow to develop new motor milestones.

However, the clinical experts indicated that patients who would be eligible for onasemnogene abeparvovec are likely to be steered to that intervention, with the exception of the potential for bridging with risdiplam until the administration of onasemnogene abeparvovec.

Patients who are most likely to benefit from risdiplam are those who might be too old for onasemnogene abeparvovec and have anatomic deformities that make nusinersen administration difficult, if not impossible. Risdiplam would essentially be the only treatment available for them.

For adolescent and adult patients with SMA, risdiplam is likely to be the only choice with some evidence of efficacy and safety. Patients with more preserved motor function at baseline will potentially see more benefit from the drug as there are more alpha motor neurons to preserve than in patients who have lost more motor function.

The clinical experts identified that patients with advanced stages of the disease, who may require invasive ventilation or have a chronic gastrointestinal condition that might hinder drug absorption, could be the least suitable for treatment with risdiplam.

Assessing Response to Treatment

Assessment of clinical progress will depend on the age of the patient, the stage of the disease, and the baseline preserved motor function.

In pre-symptomatic patients, an optimistic target is normal motor milestones development, and bulbar and respiratory function.

In patients with infantile-onset or type I SMA, achievement of independent sitting and independent breathing during daytime, the prevention of scoliosis, and functional safe swallowing is considered a good treatment response.

In patients with later-onset SMA who do not walk or type II SMA, the prevention of further deterioration and the prevention of scoliosis would be beneficial. Maintaining baseline motor function is a minimum expectation.

In patients with later-onset SMA who are able to walk or type III SMA patients, maintaining the ability to walk and potentially increasing the distance travelled in the 6-minute walk test (6MWT) would be considered a good treatment response.

Patients with adult-onset SMA or SMA type IV are unlikely to lose motor function of sufficient magnitude that would lead to loss of ambulation or disability. However, clinical experts pointed out that some patients with manual or labouring jobs may lose meaningful employment. The clinical experts do not believe that there exists evidence of the efficacy of disease-modifying therapy in this patient population.

The use of various motor function measurement tools to monitor patient progress is largely centre-dependent. Commonly, the HINE Section 2 is used for infantile-onset SMA, the HFMSE and Upper Limb Module (ULM) for non-ambulatory later-onset patients, and the HFMSE and 6MWT in ambulatory patients. However, the clinical experts have pointed out that these motor function measurement tools are not part of the common clinical practice and are usually only done as part of treatment application or reimbursement requirement. The CHOP INTEND tool was considered a research tool that requires training and resources to administer as part of the usual clinical practice. There has also been discussion regarding the utility of the 6MWT considering some identified limitations of this measuring tool, and that current assessment of motor function in ambulatory adults is largely based on self-reporting. Moreover, the expert committee pointed out that HRQoL measures may be more relevant to patients than motor function testing tools. In addition, clinical experts pointed out that most of these assessment tools require access to specialty clinics and trained staff.

Assessment of response is dependent on age and disease stage. Patients with newly diagnosed infantile-onset SMA who are treated with risdiplam are likely to start showing a response within 6 months. Adult ambulatory patients may not show the benefit of treatment for 1 year to 2 years, since disease progress is much slower in older age groups. Typically, infants are seen every 3 months until 2 years of age, then every 6 months until 4 years to 5 years of age, then annually after 5 years of age. The clinical experts suggested an assessment schedule that would mirror the typical clinical approach would be appropriate.

Discontinuing Treatment

Usual treatment discontinuation decisions would follow a risk-benefit analysis when AEs or tolerability issues are observed. Clear disease progression despite adequate treatment would be a potential reason for treatment discontinuation. However, the clinical experts emphasized that determination of disease progression should be considered carefully and not be based on measures that are single event and non-reproducible, and can potentially be due to processes other than disease progression. Any such measure would need to be objectively reproduced on at least 2 events over a 3-month period, excluding any possible acute or reversible cause.

Prescribing Conditions

Management of pediatric patients with SMA is commonly handled by a pediatric neurologist or a pediatric neuromuscular specialist within an interdisciplinary team. Adult patients with SMA are commonly managed by an adult neurologist or adult neuromuscular specialist. Treatment with risdiplam should be prescribed and administered by a clinician with experience in managing patients with SMA. For patients who are remote to specialty clinics, a hybrid model can be developed, where initial diagnosis and treatment are done at the specialty clinic, and subsequent monitoring (with prescription still coming from a specialist) is done mainly at local outpatient clinics, with periodic monitoring at the specialty clinic. Virtual visits can be considered for monitoring post-therapy initiation.

Clinician Group Input

This section was prepared by CADTH staff based on the input provided by clinician groups.

Clinician Input was provided by the NMD4C. The group is a pan-Canadian network launched in 2020 that aims to bring together clinical, scientific, technical, and patient expertise in neuromuscular disease. NMD4C is funded by the Canadian Institutes of Health Research and MDC and builds on the former neuromuscular network CAN-NMD as well as existing initiatives such as the Canadian Neuromuscular Disease Registry (CNDR) and the Canadian Pediatric Neuromuscular Group.

NMD4C noted that their mission is to improve the care, research, and treatment of neuromuscular diseases for all Canadians. They hope to be a comprehensive, inclusive, open, and enduring network through which Canadian stakeholders can share their expertise and data while also collaborating on joint activities and research. The main goals for NMD4C include formalizing and sustaining a network of neuromuscular disease stakeholders, training and educating stakeholders (clinicians, scientists, and patient advocates), improving the standard of care of neuromuscular disease patients, enabling access to therapies across Canada, and strengthening biomedical and clinical infrastructure to build research capacity in Canada.

NMD4C provided input from clinicians with experience treating SMA and with risdiplam and who are familiar with the data from the clinical trials on treatment of SMA, and specifically for risdiplam. Input was provided by neurologists from Ottawa, Montreal, Toronto, London, Vancouver, Hamilton, Calgary, and Saskatoon and included the following institutions: the CHEO Research Institute, the Montreal Neurological Institute-Hospital, Ottawa Hospital Research Institute, The Hospital for Sick Children, Children’s Hospital–London Health Sciences Centre, BC Children’s Hospital Research Institute, McMaster Children’s Hospital, Alberta Children’s Hospital, Stan Cassidy Centre for Rehabilitation, and the University of Saskatchewan. NMD4C notes that the expert clinicians are involved in clinical and observational research, clinical guidelines development, and health technology assessment.

Unmet Needs

NMD4C notes that nusinersen has been recommended for reimbursement by the CADTH Canadian Drug Expert Committee and most children with SMA are receiving nusinersen as part of their treatment regimen. The clinicians noted that nusinersen is injected into the spinal fluid every 4 months after an initial 4 loading doses, which occur during the first 2 months of treatment. Additionally, the clinician group noted that this procedure is typically done under sedation at an experienced pediatric centre. NMD4C also noted that patients receiving nusinersen showed improvement in motor function, reducing the need for assisted ventilation, and greatly extending survival when compared to sham injection; it is considered to be a significant advancement in treating patients with SMA. However, while nusinersen has significantly changed the standard of care for children and youth with SMA, the clinicians note that this treatment Is not suitable for all patients with SMA. These include patients who have ever walked (any child with type III disease) who are excluded from reimbursement, additionally, in the clinical trials for nusinersen; not all patients with type I SMA and type II SMA responded to treatment. NMD4C added that the use of nusinersen in some patients is also limited by the invasive intrathecal route of administration and that this treatment is more challenging for patients who have severe scoliosis or who have undergone a spinal fusion. The clinicians added that there are a number of children who do not tolerate lumbar puncture procedures due to respiratory fragility or poor access to the spine, and some patients experience side effects from repeated lumbar punctures. Access was also noted as an issue since specialists and facilities who can administer the treatments are not universal and barriers such as long-distance travel exist.

NMD4C added that a very limited number of Canadian children have had access to onasemnogene abeparvovec (Zolgensma). The clinicians noted that onasemnogene abeparvovec is a 1-time IV infusion treatment for infants with SMA type I.

NMD4C added that for patients with early-onset SMA, preserving motor neurons, improving survival, improving motor function, delaying or alleviating the need for assisted ventilation, delaying or alleviating the loss of the ability to speak, and delaying other secondary complications (such as failure to thrive, scoliosis, recurrent pulmonary infections, and so forth), and reducing burden on caregivers are goals that new treatments would address. The clinicians noted that while the therapies discussed previously produce the missing SMN protein, there remains a need for more convenient and less invasive treatments for adults with SMA. Additionally, therapies that improve muscle functions are needed and would play a pivotal role in patient care.

The clinicians noted that for individuals with late-onset SMA, treatment goals are to maintain the current level of motor function (prevent further loss of motor function) and strength, achieve disease stabilization (prevent disease progression, including avoidance of need for ventilation), promote independence, and improve overall HRQoL.

NMD4C noted that risdiplam is important for SMA patients with contraindications for nusinersen or Zolgensma. The following unmet needs were discussed by NMD4C:

• Risdiplam is an oral therapy. Oral therapy is required for children (< 18 years) who are unable to receive intrathecal nusinersen by simple lumbar puncture due to severe scoliosis or post-spinal fusion.

• Adult SMA patients (≥ 18 years) have a high unmet need for treatment. The clinicians recognize that the evidence available is limited and suggest that provinces support the capture of outcomes data of SMA patients in the CNDR to address uncertainty for this population.

NMD4C clinicians noted that SMA patients have a very high burden of illness and health care resource use; having a choice in therapy that is based on strong evidence and matches their preferences and values should be the goal. Risdiplam was noted as an additional option to treat SMA.

Place in Therapy

NMD4C noted that an oral route of administration has advantages for both patients and caregivers as it does not require specialized procedures to administer treatment and allows patients to take the medication at home, ultimately avoiding frequent travel and reducing health care utilization costs. Additionally, the clinicians added that invasive procedures such as intrathecal injections carry risk and are not an option for all patients due to underlying scoliosis. NMD4C added that they expect risdiplam to be used as an option in the first-line treatment of early-onset type I SMA; however, the treatment could also be used in patients with previous exposure to SMN-splicing agents and could be used as a second-line treatment of patients who have not responded or progressed on first-line treatment such as nusinersen. Patients with later-onset SMA (type II and type III) and patients who are not eligible for nusinersen due to age or physical ability would be considered, though younger patients are expected to achieve the most benefit.

NMD4C noted that Zolgensma may become the preferred initial treatment for babies (under the age of 2 years) if it is approved and reimbursed in Canada. The clinicians added that this is because it is a single IV infusion, and that the systemic (IV) delivery route allows for wider CNS and peripheral tissue distribution, which is important for infants with type I SMA as high levels of SMN protein are required very early in life in peripheral tissues.

Due to the lack of head-to-head comparison, the clinicians noted that there is uncertainty about whether children with SMA would derive more benefit from nusinersen or from risdiplam. Additionally, the clinicians added that risdiplam may be preferred over nusinersen in circumstances where the intrathecal administration (of nusinersen) is not possible. Furthermore, the clinicians noted that risdiplam has an oral route of administration and allows for distribution of SMN protein in both the CNS and peripheral tissues, which may make it the preferred option for patients where stabilization of the disease would mean retention of a vital motor function, the avoidance of ventilator dependency, the continued ability to speak and swallow, and survival.

With respect to sequencing, NMD4C noted that risdiplam would potentially be first-line therapy for patients who are not eligible for Zolgensma or nusinersen as per the drug label or reimbursement criteria. The clinicians added that risdiplam may be preferred when patients cannot receive nusinersen for safety reasons such as scoliosis, allergies, or bleeding disorders, when children experience adverse effects to nusinersen, and where there is lack of access to nusinersen, based on patient preferences and in individuals who have antibodies to AAV9. Risdiplam can also be considered second-line therapy for patients where first-line treatment has been unsuccessful or worn off. The clinicians added that nusinersen has an intrathecal route of administration, which may limit its effects on the CNS because it may not distribute equally along the spinal cord. The clinicians added that it is hypothesized that nusinersen does not reach motor neurons in the upper spinal cord as effectively as motor neurons in the lower spinal cord (where the injection site is at the lumbar spine) and, therefore, bulbar functions such as swallowing are not adequately treated.

Patient Population

NMD4C noted that generally all 3 treatments (nusinersen, Zolgensma, and risdiplam) work best the younger the children are and the earlier the treatment begins. The clinicians added that patients benefiting the most from risdiplam are likely not different from those benefiting the most from nusinersen. The unmet need, according to the clinicians, is for children who cannot receive nusinersen, for reasons discussed earlier, and for adult patients.

NMD4C noted that for many adults living with SMA, the duration of the disease is associated with progressive losses of motor function, deterioration in health, and reduced independence. The clinicians noted that stabilization of the disease can lead to the retention of vital motor function(s), the avoidance of ventilator dependency, the continued ability to speak and swallow, and survival.

NMD4C noted that patients can be identified before initiating treatment through genetic testing and confirmed by the absence of normal copies of the SMN1 gene. The clinicians added that a major disease modifier is the number of SMN2 gene copies, where fewer copies are associated with earlier-onset and more severe SMA. The clinicians also added that identifying infants through newborn screening will allow for maximal therapeutic benefit through the administration of treatment pre-symptomatically.

The clinicians noted that children and adults with SMA are typically followed and managed by specialized clinical teams, and genetic testing to confirm diagnosis and SMN2 gene copies can be obtained if not already available. They added that patients with the “SMA 0” form of 5q SMA are unlikely to benefit and patients with other genetic types of SMA, unrelated to SMN deficiency, are also unlikely to benefit.

NMD4C added that patients with fewer copies of the SMN2 gene would be associated with earlier-onset and more severe SMA. The clinicians added that most patients will benefit from the drug, possibly apart from severely affected cases who are on permanent ventilation. In general, the youngest patients would benefit more.

The clinicians added that children who have been identified through newborn screening and treated with Zolgensma or nusinersen in the pre-symptomatic stage may show normal development, in which case they may not require treatment with risdiplam.

Assessing Response to Treatment

NMD4C noted that outcome tests vary according to age and functional state. The clinicians added that tests may include lung function (FVC), RULM, and 6MWT. NMD4C added that in small children, there are tests that have different sensitivity for the achievement of motor milestones. The Revised Hammersmith Scale was noted as a scale specifically designed for outcome measures for people affected by SMA; however, that scale might lack sensitivity and has a “floor effect,” which may make the 32-item MFM-32 a preferred measure for younger individuals aged 2 years to 5 years. The clinicians noted that MFM-32 may also be preferred in non-ambulant individuals with type II SMA or type III SMA, aged 2 years to 25 years. In Canada, the clinicians noted that the Revised Hammersmith Scale is widely used and accepted as the tool to measure motor function. The group added that in small children, tests will also vary based on whether they achieve motor milestones. All these measures require trained practitioners and would be considered outside the bounds of a traditional clinic visit.

When assessing outcomes, clinician experts generally agree that patient-reported outcome instruments may be useful, although there are no internationally validated and agreed-upon patient-reported outcome instruments to date and data are being developed to inform the selection of such a measure.

NMD4C added that in 2020, Canadian experts in adults with SMA carried out a modified Delphi process exercise to determine a consensus-based recommendation for outcome measures that can be used in adults with SMA at different functional stages.

A clinically meaningful response, as per the clinicians who provided input, would be the achievement of motor milestones, ventilation-free survival, and the improvement or stabilization of motor function in children. For adults, the clinicians noted the stabilization of motor and respiratory function, less disability with the maintenance of independence, and fewer hospitalizations would be considered a clinically meaningful response. The clinicians added that maintaining the ability to speak and avoiding the need for ventilation support have profound effects on a patient’s quality of life, autonomy, and ability to maintain vocational and social roles.

The clinicians noted that with new therapies for SMA introduced to clinical practice, it is important to monitor effectiveness and collect evidence to help determine which treatment option is best suited for each patient. They also noted that quantitative outcome measures require specially trained practitioners, are time consuming, and are not currently covered with provincial and territorial funding programs.

NMD4C provided excerpts from a letter to the provincial government outlining its concerns and recommendations for an alternative timeline for outcome measurements in patients receiving SMA therapies. These recommendations are noted as follows.

Frequency of Monitoring

The natural history of SMA is a slowly progressive functional decline. Patients’ function decreases by a statistically significant amount in a matter of years, not months. As such, any decline in outcome measure scores in the span of months is more likely to be due to inter- or intra-examiner variability (i.e., the normal statistical range of error) or variability in patient effort (i.e., SMA patients having off days and good days just like anyone else) than due to a true effect of the therapy. This potential artificial decline in outcome measure scores could lead to inappropriate termination of therapy provision.

Human Resource Strain and Patient Burden

Motor outcome assessments are time consuming and personnel-intensive, requiring significant time from physical therapists and/or respiratory therapists with specialized training. As such, there are ethical concerns around mandating that these scarce resources be used to provide frequent monitoring for largely stable patients for whom such frequent monitoring is not clinically relevant or evidence-based. Frequent clinic visits can also be fatiguing, time consuming, and costly for patients.

Collection of Real-World Evidence and Proposition for Appropriate Monitoring

The CNDR has a clear plan to assess the effectiveness of novel therapies in a broad population. The CNDR’s SMA working group of expert investigators has evaluated specific outcome assessments by way of a Delphi model to determine best practices (a manuscript is in preparation). They recently agreed that the best timing for outcome measurements is at baseline, 6 months later, then annually thereafter. This timing is appropriate for clinical monitoring, congruent with available evidence in treated adults, and does not put undue burden on provincial and territorial clinical resources or on patients and families.

Discontinuing Treatment

NMD4C notes that factors to consider when deciding to discontinue treatment (and switch to other SMN-splicing modulating treatments) include accelerated deterioration in clinical status while on risdiplam for 12 months, allergic reaction and critical SAEs, and patients with primary retinal disease who may need special protocol and/or assessments. The clinicians noted that a variety of biomarkers are being investigated that may help to predict the variability of response to treatment and may allow for a more individualized treatment pathway.

Prescribing Conditions

NMD4C noted that risdiplam would offer the convenience of oral administration, which can take place at home. The clinicians added that it may also be appropriate when a patient cannot receive nusinersen for safety reasons, when children experience adverse effects to nusinersen, when there is no access to nusinersen within a given jurisdiction, and based on patient preferences, as well as when individuals have contraindications to receiving Zolgensma, such as antibodies for AAV9.

NMD4C noted that neurologists, pediatric neurologists, and physiatrists would be needed to diagnose, treat, and monitor patients who might receive risdiplam.

Additional Considerations

NMD4C added that in Canada, the CNDR has recruited 4,000 patients with various neuromuscular diseases since its launch in 2010. The registry will be further developed to capture information that is useful for the diseases in question and will continue to support academic- and industry-led research, including quality of life, burden of illness, and preference studies.

Drug Program Input

The drug programs provide input on each drug being reviewed through CADTH’s reimbursement review processes by identifying issues that may impact their ability to implement a recommendation. The implementation questions and corresponding responses from the clinical experts consulted by CADTH are summarized in Table 4.

Clinical Evidence

The clinical evidence included in the review of risdiplam is presented in 3 sections. The first section, the systematic review, includes pivotal studies provided in the sponsor’s submission to CADTH and Health Canada, as well as those studies that were selected according to an a priori protocol. The second section includes indirect evidence from the sponsor and indirect evidence selected from the literature that met the selection criteria specified in the review. The third section includes sponsor-submitted long-term extension studies and additional relevant studies that were considered to address important gaps in the evidence included in the systematic review.

Systematic Review (Pivotal and Protocol-Selected Studies)

Objectives

To perform a systematic review of the beneficial and harmful effects of risdiplam for the treatment of SMA, administered once daily as an oral solution at a dose of 0.20 mg/kg for patients 2 months of age and up to 2 years, 0.25 mg/kg for patients 2 years of age or older who weigh less than 20 kg, and 5 mg for patients 2 years of age or older who weigh 20 kg or more

Methods

Studies selected for inclusion in the systematic review include pivotal studies provided in the sponsor’s submission to CADTH and Health Canada, as well as those meeting the selection criteria presented in Table 5. Outcomes included in the CADTH review protocol reflect outcomes considered to be important to patients, clinicians, and drug plans.

The literature search was performed by an information specialist using a peer-reviewed search strategy. The literature search for clinical studies was performed by an information specialist using a peer-reviewed search strategy according to the PRESS Peer Review of Electronic Search Strategies checklist.²⁷

Published literature was identified by searching the following bibliographic databases: MEDLINE All (1946‒) via Ovid and Embase (1974‒) via Ovid. The search strategy comprised both controlled vocabulary, such as the US National Library of Medicine’s MeSH (Medical Subject Headings), and keywords. The main search concepts were Evrysdi (risdiplam). Clinical trials registries were searched: the US National Institutes of Health’s clinicaltrials.gov, the WHO’s International Clinical Trials Registry Platform (ICTRP) search portal, Health Canada’s Clinical Trials Database, and the European Union Clinical Trials Register.

No filters were applied to limit the retrieval by study type. Retrieval was not limited by publication date or by language. Conference abstracts were excluded from the search results. See Appendix 1 for the detailed search strategies.

The initial search was completed on December 17, 2020. Regular alerts updated the search until the meeting of the CADTH Canadian Drug Expert Committee was held on April 21, 2021.

Grey literature (literature that is not commercially published) was identified by searching relevant websites from the Grey Matters: A Practical Tool For Searching Health-Related Grey Literature checklist.²⁸ Included in this search were the websites of regulatory agencies (the FDA and EMA). Google was used to search for additional internet-based materials. See Appendix 1 for more information on the grey literature search strategy.

These searches were supplemented by reviewing bibliographies of key papers and through contacts with appropriate experts. In addition, the manufacturer of the drug was provided information regarding unpublished studies.

Two CADTH clinical reviewers independently selected studies for inclusion in the review based on titles and abstracts, according to the predetermined protocol. Full-text articles of all citations considered potentially relevant by at least 1 reviewer were acquired. Reviewers independently made the final selection of studies to be included in the review, and differences were resolved through discussion.

In addition, a focused literature search for NMAs dealing with SMA was run in MEDLINE All (1946‒) and Embase (1974‒) on December 17, 2020. No limits were applied to the search.

Findings From the Literature

Two studies were included in the systematic review (Figure 3). The included studies are summarized in Table 6. A list of excluded studies is presented in Appendix 2.

Figure 3: Flow Diagram for Inclusion and Exclusion of Studies

Description of Studies

One single-arm uncontrolled trial (FIREFISH) and 1 randomized controlled trial (SUNFISH) were included in this CADTH reimbursement review.

FIREFISH (also known as Study BP39056) is an ongoing, open-label, single-arm, clinical trial to investigate the safety, tolerability, PK, PD, and efficacy of risdiplam in patients with type I SMA. The study was performed at sites in various countries internationally but did not include sites in Canada. FIREFISH was conducted in 2 parts: a phase II exploratory dose-finding part (Part 1) and a phase III confirmatory part (Part 2). Enrolment for Part 1 started on September 23, 2016. Upon establishing the therapeutic dose, the first patient in Part 2 was enrolled on March 13, 2018. SUNFISH (also known as Study BP39055) is an ongoing, randomized, double-blind, placebo-controlled trial to investigate the efficacy, safety, tolerability, PK, and PD of risdiplam in pediatric and adult patients with type II SMA and type III SMA. The study was performed at sites in various countries internationally, including Canada.

In a similar manner to FIREFISH, SUNFISH was conducted in 2 parts: a phase II exploratory dose-finding part (Part 1) and a phase III confirmatory part (Part 2). Enrolment for Part 1 started on October 19, 2016. Upon establishing the therapeutic dose, the first patient in Part 2 was enrolled on October 9, 2017. The 2 parts ran separately and no patients from Part 1 were moved to Part 2. Considering the confirmatory phase III nature of Part 2 and the dose-finding phase II nature of Part 1, only details and results of Part 2 are presented in this section, while a summary of Part 1 is presented in the Other Relevant Evidence section.

FIREFISH Part 2 is an ongoing single-arm study that aimed to assess the efficacy of risdiplam (dose based on age) on patients with infantile-onset SMA aged between 28 days and 3 months with 2 copies of the SMN2 gene. The study used the term “type I SMA” to indicate the diagnosis of the included patients. Patients were treatment-naive, with no invasive ventilation and an arterial oxygen saturation level greater than 94%. The study’s primary outcome was to assess the proportion of infants who are able to achieve the motor milestone of sitting independently for 5 seconds after 12 months of treatment. The study duration is 24 months.

SUNFISH Part 2 is an ongoing, randomized, double-blind, placebo-controlled trial for a 12-month period, followed by a second 12-month period where patients receiving placebo switched to risdiplam in a blinded manner and all patients continued risdiplam until month 24. SUNFISH Part 2 aimed to assess the efficacy of risdiplam compared to placebo over a 12-month treatment period, and to report on the overall efficacy and safety of risdiplam over a 24-month treatment period. Risdiplam was administered orally at a dose of 5 mg once daily for patients with body weight of 20 kg or over and 0.25 mg/kg once daily for patients with body weight less than 20 kg. Matching oral placebo was administered once daily. SUNFISH Part 2 studied the effect of risdiplam in patients aged 2 years to 25 years diagnosed with SMA type II and non-ambulatory patients diagnosed with SMA type III. Patients were randomized on a 2:1 ratio to risdiplam and placebo, respectively, using an interactive response system. Randomization was stratified by age groups (2 years to 5 years, 6 years to 11 years, 12 years to 17 years, and 18 years to 25 years). The study’s primary outcome was to assess the change from baseline in the total score of the MFM-32 at 12 months of treatment. The study duration is 24 months.

Populations

Inclusion and Exclusion Criteria

FIREFISH Part 2 aimed to enrol patients with SMA who have the highest medical need and who would benefit the most from early initiation of risdiplam. This translated into patients who were aged 1 month to 7 months diagnosed with what the sponsor described as type I SMA. This was defined in FIREFISH Part 2 as patients with homozygous or heterozygous deletion or point mutation in the SMN1 gene, showing signs and symptoms attributed to type I SMA, and having 2 SMN2 gene copies. In addition, patients in FIREFISH Part 2 had to be treatment-naive with a relatively normal respiratory function. Specifically, patients who required invasive ventilation were excluded along with patients with awake hypoxemia (oxygen saturation < 95%) regardless of ventilation use. Further exclusion criteria were placed that would exclude patients with potential concomitant conditions or therapeutic uses that, in the opinion of the sponsor, would have interfered with an unbiased assessment of the effects of risdiplam.

SUNFISH Part 2 aimed to include all non-ambulatory patients between the ages of 2 years and 25 years who were diagnosed with SMA type II or SMA type III. Patients had to have a minimum motor function requirement, including a RULM item A of 2 or more and the ability to sit independently. Patients were never to have had a disease-modifying therapy, were not to have been hospitalized in the 2 months before screening due to a pulmonary event, and were not to have had a tracheostomy. Further exclusion criteria were placed that would exclude patients with major illnesses within 1 month, unstable physiology, and a history of ophthalmic diseases as determined by the investigator.

Baseline Characteristics

A total of 41 patients were enrolled in FIREFISH Part 2 and were given risdiplam. These patients had an average age of 5.2 (SD = 1.47) months; 53.7% were females and 53.7% were White. All of the included patients had 2 SMN2 gene copies and the onset of symptoms was reported at a mean age of 1.64 (SD = 0.70) months. Average disease duration in FIREFISH Part 2 was reported at 3.59 (SD = 1.35) months. At baseline, |||| of the patients were able to control their head upright, while |||| did not demonstrate any motor milestone achievement. Of the enrolled patients, 70.7% did not require respiratory support while the rest may have used a non-invasive respiratory support for less than 16 hours per day. Ability to swallow was demonstrated in 97.6% of the enrolled patients at baseline. A summary of baseline characteristics of FIREFISH Part 2 is presented in Table 7.

In SUNFISH Part 2, 120 patients were randomized to risdiplam and 60 patients to placebo. The mean age of enrolled patients was 9.9 (SD = 5.8) years in the risdiplam group and 10.3 (SD = 6.1) years in the placebo group. Patients belonging to the age group of 18 years to 25 years formed the smallest age group (11.7% in risdiplam, 13.3% in placebo), followed by the age group of 12 years to 17 years (25.0% in risdiplam, 26.7% in placebo). There were similar proportions of patients in the 2 youngest age groups: 32.5% and 30.0% were in the risdiplam and placebo arms of the 6 years to 11 years age group, respectively, and 30.8% and 30.0% were in the risdiplam and placebo arms of the 2 years to 5 years age group, respectively. Most patients had 3 SMN2 gene copies (89.2% in risdiplam, 83.3% in placebo), while more than two-thirds were diagnosed as having SMA type II (70.0% in risdiplam, 73.3% in placebo). At baseline, |||| were able to stand in the risdiplam arm and |||| in the placebo arm. Discrepancies in the baseline characteristics can be observed in the proportion of patients with 3 SMN2 copies (89.2% in the risdiplam arm and 83.3% in the placebo arm), and in the |||||||||||||||||||||||||||||||||| |||||| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||. A summary of baseline characteristics of SUNFISH Part 2 is presented in Table 8.

Interventions

FIREFISH Part 2

Patients enrolled in FIREFISH Part 2 were given risdiplam orally once daily. The dose was determined based on the age of the patient as follows:

• infants older than 1 month old and younger than 3 months old at enrolment: 0.04 mg/kg

• infants 3 months old or older and younger than 5 months old at enrolment: 0.08 mg/kg

• infants 5 months old or older at enrolment: 0.2 mg/kg.

The dose was determined at the age of enrolment and did not change automatically once a patient reached 3 months or 5 months of age, with the exception of cases where a change of dose is requested by the clinical pharmacologist upon review of the PK and safety data. The PK was monitored every 2 weeks to ensure patients stayed within the targeted exposure range of a mean area under the curve from 0 hours to 24 hours at steady state (AUC_0-24,ss) of 2,000 ng*hour/mL or less.

Permitted concomitant therapy included any stable regimen therapy for a chronic condition (initiated 6 weeks before enrolment) or therapies to provide symptomatic treatment, with the exception of the following explicitly prohibited medications:

• any inhibitor of cytochrome P450, family 3, subfamily A, member 4 (CYP3A4), including, but not limited to, ketoconazole, miconazole, itraconazole, fluconazole, erythromycin, clarithromycin, ranitidine, cimetidine

• any inducer of CYP3A4, including, but not limited to, rifampicin, rifabutin, glucocorticoids, carbamazepine, phenytoin, phenobarbital, St. John’s wort

• any organic cation transporter-2 (OCT-2) and multidrug and toxin extrusion (MATE) substrates, including, but not limited to, amantadine, cimetidine, memantine, amiloride, famotidine, metformin, pindolol, ranitidine, procainamide, varenicline, acyclovir, ganciclovir, oxaliplatin, cephalexin, cephradine, fexofenadine

• any known FMO1 or FMO3 inhibitors or substrates

• any medication with known phototoxicity and retinal toxicity liabilities.

SUNFISH Part 2

Patients randomized to the risdiplam arm in SUNFISH Part 2 were given risdiplam orally once daily. The dose was determined based on the weight of the patient as follows:

• 5 mg once daily for patients with body weight of 20 kg or more

• 0.25 mg/kg once daily for patients with body weight of less than 20 kg.

Patients who started on the 0.25 mg/kg dose were adjusted according to their weight at the next study visit, if required. The treatment was to be discontinued in cases of pregnancy, lack of compliance with study requirements, in cases of ophthalmic AEs, in cases of significant elevation of alanine transaminase or aspartate transaminase (> 3x upper limit) that is associated with a twofold increase in bilirubin, and other SAEs that suggest dosing should be halted.

Permitted concomitant therapy included any stable regimen therapy (initiated 6 weeks before enrolment) for a chronic condition or therapies to provide symptomatic treatment, with the exception of the following explicitly prohibited medications:

• any medication that was part of the exclusion criteria of the study (e.g., nusinersen)

• any OCT-2 and MATE substrates (e.g., amantadine, cimetidine, memantine, amiloride, famotidine, metformin, pindolol, ranitidine, procainamide, varenicline, acyclovir, ganciclovir, oxaliplatin, cephalexin, cephradine, fexofenadine)

• deferoxamine, topiramate, latanoprost, niacin, rosiglitazone, tamoxifen, canthaxanthin, sildenafil, interferon or any other drugs known to cause retinal toxicity, including the chronic use of minocycline.

Outcomes

A list of efficacy end points identified in the CADTH review protocol that were assessed in the clinical trials included in this review is provided in Table 9. These end points are further summarized as follows. A detailed discussion and critical appraisal of the outcome measures is provided in Appendix 3.

FIREFISH Part 2

Sitting Without Support After 12 Months of Treatment and Other BSID-III Gross Motor Scale–Related Outcomes

The BSID-III gross motor scale is 1 of 2 subdomains of the Motor scale in the BSID-III tool. The Motor scale itself is 1 of 5 scales that constitute the BSID-III tool. The BSID-III is a norm-referenced instrument designed to measure the developmental status of children aged 1 month to 42 months. The BSID-III gross motor scale includes a total of 77 items that assess static positioning, dynamic movement, quality of movement, balance, motor planning, and perceptual-motor integration. While there is no established MID, based on the classical phenotypes of the SMA population, type I SMA patients never sit independently while type II SMA patients never walk independently.^29-31

The sponsor administered a modified version of the BSID-III, where the gross motor scale consisted of 72 items scored at 0 (unable to perform the activity) or 1 (criteria for item achieved). The total raw score was calculated by summing the item scores to give a maximum possible score of 72. If an individual item was missing or “cannot test” was recorded, the item score was set to 0.

The outcome of sitting without support after 12 months of treatment was the primary end point of the confirmatory Part 2 of the study. Sitting was defined based on the BSID-III gross motor subset item 22, outlined as “sits alone without support for at least 5 seconds.”

Assessment of this outcome, and all other BSID-III gross motor scale–related outcomes, was video-recorded at the study site and responders were determined at the site by the clinical evaluator, and centrally by 2 independent readers. The 2 independent central readers only reviewed and scored items 22, 26, 30, 40, and 42. The assessment of the 2 independent readers will be used for the primary analysis. In cases where there was disagreement between the 2 independent reviewers, they conducted a second assessment of the disagreed-upon result; if disagreement persisted after the second round of assessment, then the patient would not have been considered to have achieved the outcome.

CHOP INTEND Score and Other CHOP INTEND–Related Motor Milestone Outcomes

CHOP INTEND was developed in infants with SMA type I to measure motor function in children with neuromuscular disorders presenting in infancy and possessing an infant’s repertoire of motor skills. The instrument consists of 16 items rated on a uniform scale from 0 to 4, corresponding to no response (0), minimal (1), partial (2), nearly full (3), and complete (4) response levels. The 16 items include spontaneous movement (upper extremity), spontaneous movement (lower extremity), hand grip, head in midline with visual simulation, hip adductors, rolling (elicited from legs), rolling (elicited from arms), shoulder and elbow flexion and horizontal abduction, shoulder flexion and elbow flexion, knee extension, hip flexion and foot dorsiflexion, head control, elbow flexion, neck flexion, head and neck extension, and spinal incurvation. The maximum total score is 64; higher scores indicate more advanced motor development. No published estimated MID for CHOP INTEND was identified.^32,33

The sponsor used the CHOP INTEND tool in 3 outcomes: the proportion of patients who achieve a score of 40 or higher in the CHOP INTEND at month 12, the proportion of patients who achieve an increase of at least 4 points in their CHOP INTEND score from baseline at month 12, and the proportion of patients who achieve head control as assessed by item 12 of the CHOP INTEND.

HINE Section 2 Score and Other HINE Section 2–Related Motor Milestone Outcomes

HINE Section 2 is composed of 8 milestones: head control, sitting, voluntary grasp, ability to kick, rolling, crawling, standing, and walking. Each milestone has 3 to 5 possible descriptive ratings, ranging from not performing the task at all to fully demonstrating the milestone; the motor milestones are age-dependent and are not intended to produce a total score. Although the original HINE developers did not define a quantitative scoring system for section 2, scores for each milestone were obtained by assigning a value of 0 to the absence of the activity and adding 1 point for each incremental rating. The total score is calculated by summing the item scores to give a maximum possible score of 26. If an individual item was missing or if “cannot test” was recorded, the item score was set to 0.³⁴

The sponsor defined an improvement in a motor milestone as at least a 2-point increase in ability to kick (or maximal score) or a 1-point increase in head control, rolling, sitting, crawling, standing, or walking. Worsening is similarly defined as a 2-point decrease in ability to kick (or lowest score) or a 1-point decrease in head control, rolling, sitting, crawling, standing, or walking. Voluntary grasp is excluded from the definition. Based on this, an infant was defined as a responder if more motor milestones showed improvement than showed worsening; infants who died or were withdrawn were classified as nonresponders. Infants with a totally missing HINE Section 2 assessment at month 12 or month 24 were classified as nonresponders.

Other outcomes that the sponsor used based on the HINE Section 2 are the proportion of patients able to support weight or stand with support as assessed by the HINE Section 2 at month 12, and the proportion of patients able to bounce while assessing the walking item of the HINE Section 2 at month 12.

Survival, Ventilation-Free Survival, and Respiratory-Related Outcomes

Time to death, time to death or permanent ventilation, and time to permanent ventilation was presented graphically using the Kaplan-Meier curves. Permanent ventilation is defined as 16 hours or more of non-invasive ventilation (e.g., BiPAP) per day or intubation for more than 21 consecutive days in the absence of, or following the resolution of, an acute reversible event; or tracheostomy. A permanent ventilation event was determined through an adjudication committee. Time to event was defined as the time in months from the date of enrolment in the study until the date of the event.

Nutrition

One of the secondary outcomes was the proportion of patients with the ability to feed orally at month 12. A combination of oral and tube feeding was also classified as oral feeding. Infants who had not maintained the ability to feed orally if achieved earlier, or had been withdrawn, or died, were classified as nonresponders for the analysis. Infants with a missing assessment at a visit were also classified as nonresponders.

Health Care Utilization

Hospitalizations were defined as any hospital admission for 2 or more days for any reason that is not due to the study requirement.

The number of hospitalizations per patient-year was defined as the number of hospitalizations observed by month 12 (or month 24) divided by the total patient-years at risk, where the total patient-years at risk were equal to the sum across all patients of the time intervals (in years) between the start of study therapy and the earliest date of study withdrawal or completion of 12 months (or 24 months) of treatment.

Infant Toddler Quality of Life Questionnaire–47-Item Short Form

The Infant Toddler Quality of Life Questionnaire (ITQOL) is a 97-item, caregiver-reported, generic profile measure used to assess health status and HRQoL for infants and children aged 2 months to 5 years. The ITQOL-SF47 is a shortened 47-item version that was used in FIREFISH. It uses a 5-point Likert-style response from “all of the time” to “none of the time” to evaluate physical function (6 items), growth and development (5 items), bodily pain (2 items), temperament and moods (6 items), behaviour (12 items), general health perceptions (5 items), parent emotional health (4 items), and parent time limitations (4 items). The ITQOL also includes single-item questions specific to overall health, change in health, and family cohesion. Higher scores indicate better HRQoL, and raw scale scores are transformed to a scale from 0 to 100. No literature was identified that assessed the ITQOL-SF47 for validity, reliability, or responsiveness in patients with SMA. No MID was identified in populations with SMA.

SUNFISH Part 2

Change From Baseline in MFM-32 Score at Month 12 and Other MFM-32–Related Outcomes

The MFM-32 is used to assess the motor function of patients with neuromuscular disease. The 32 items are categorized in 3 domains (Ds): D1 (13 items) evaluates standing, transfers, and ambulation; D2 (12 items) measures proximal and axial function; and D3 (7 items) assesses distal function.^35,36 A patient’s maximal ability without assistance is scored on a 4-point Likert scale: 0 = movement not initiated or starting position not maintained; 1 = exercise partially completed; 2 = exercise completed with compensations, slowness, or obvious clumsiness; and 3 = exercise completed with standard pattern. Subscores can be calculated as a percentage of the maximum possible scores for each domain.³⁶ The raw score (sum of the 32 items, ranging from 0 to 96) is divided by 96 (maximum score) and multiplied by 100 to produce a final score that can be interpreted as a “percentage of normal function;” a lower score indicates lower motor function.^37,38 The sponsor indicated that a change of 3 or more points in the MFM-32 may translate into either the acquisition of a new function or the improvement in performance of several functions. As such, the first secondary outcome tested was the proportion of patients with a change from a baseline MFM-32 total score of 3 or more at month 12. Furthermore, the sponsor has presented arguments that — considering the deteriorating nature of the disease — stabilization and the lack of further deterioration may also be an important outcome. This was measured in a secondary outcome that was outside the statistical testing hierarchy: the proportion of patients with a change from a baseline MFM-32 total score of 3 or more at month 12.

In addition, the sponsor reported on the change from baseline in each of the 3 domains individually, the combined score of D1 + D2, and the combined score of D2 + D3.

Change From Baseline in the Total Score of the RULM at Month 12 and Other RULM–Related Outcomes

The original ULM was designed to capture upper limb function in non-ambulatory SMA patients, especially in young children, and was previously validated in this population.³⁹ Due to ceiling effects, it was revised and renamed the RULM. It consists of 19 items that are graded on 3-point scales. With the exception of 1 activity with a binary score, the possible scores are as follows: 0 (unable), 1 (able, with modification), and 2 (able, no difficulty), giving a maximum total score of 37 where a higher score indicates greater function. The patient chooses 1 arm with which to perform the tasks. Stolte et al. calculated MIDs based on the standard error of measurement (SEM) with a range of 1.2 to 2.7 in patients with SMA type II, and a range of 2.7 to 5.9 for patients with SMA type III.⁴⁰

Change from baseline was a secondary outcome. Other related outcomes included the proportion of patients with a change of 2 or more, and the proportion of patients with a change of 0 or more, both of which were outside the statistical testing hierarchy.

Change From Baseline in Total Score of the HFMSE at Month 12 and Other HFMSE-Related Outcomes

The HFMSE is intended for use in patients with SMA type II and SMA type III and captures higher functioning skills. It consists of 33 activities that can be scored as follows: 0 for unable to perform, 1 for performs with modification and/or adaptation, and 2 for performs without modification. The item scores are summed to give a total score with a maximum of 66. The higher the total score, the greater the patient’s motor functioning.⁴¹ An MID calculated based on SEM shows a range of 0.5 to 1.2 in patients with type II SMA, and a range of 2.7 to 5.9 in patients with type III SMA.⁴⁰

Change from baseline was a secondary outcome. Other related outcomes included the proportion of patients with a change of 2 points or more, and the proportion of patients with a change of 0 or more, both of which were outside the statistical testing hierarchy.

Respiratory-Related Outcomes

Several respiratory-related outcomes were reported in SUNFISH. The most relevant of these is the FVC, which is the total volume that can be exhaled after inhaling maximally, and the forced expiratory volume in 1 second (FEV₁), which is the volume forcefully exhaled in the first second of the FVC test. Change from baseline in best predicted FVC was a secondary outcome; other respiratory-related outcomes included the FEV₁. FVC and FEV₁ were obtained from spirometry tests that were administered to patients who were 6 years to 25 years of age at screening.

Spinal Muscular Atrophy Independence Scale

The SMAIS was developed by the sponsor for the SUNFISH study as a patient-reported outcome tool that is specific to SMA type II patients and non-ambulatory SMA type III patients. The SMAIS consists of 29 items and measures the level of independence a patient requires to perform daily activities.¹⁴ It has 2 versions with identical items: 1 that is self-reported for patients older than the age of 12 and another that is caregiver-reported for patients older than 2 years of age. Twenty-two of the 29 items cover activities that use the upper limbs, such as personal hygiene, eating, drinking, and writing. The other 7 items focus on activities that rely on lower limb and proximal and/or axial movements such as mobility and completing chores. Each item is scored on a 3-point scale where 0 = unable to perform task at all without help, 1 = needs a lot to a moderate amount of help, and 2 = needs little to no help. A higher overall score indicates greater independence. An MID was calculated using anchor-based analyses for the ULM of the SMAIS scale (SMAIS-ULM); it was estimated that a 1-point to 5-point change on the SMAIS-ULM would be clinically meaningful.

Change from baseline in caregiver-reported SMAIS total score was a secondary outcome.

Clinical Global Impression–Change

The CGI-C is used to evaluate the change in a patient’s global health from baseline.³⁵ It is a single-item, 7-point Likert scale with the following options: 1 = very much improved; 2 = much improved; 3 = minimally improved; 4 = no change; 5 = minimally worse; 6 = much worse; and 7 = very much worse. No literature was identified that assessed the CGI-C for validity, reliability, or responsiveness in patients with SMA. No MID information was identified in patients with SMA.

The proportion of patients rated as improved at month 12 was a secondary outcome.

Statistical Analysis

FIREFISH Part 2

FIREFISH Part 2 was a single-arm, open-label, uncontrolled, phase III clinical trial.

The primary outcome of Part 2 was the proportion of infants who were able to sit without support at 12 months of treatment; this outcome was tested against a predefined performance criterion of 5%. This 5% threshold was chosen based on the history of type I SMA, in which, as part of the classification, infants never achieve the motor milestone of sitting without support.

Based on the predefined performance criterion of 5%, an assumption was made that the true proportion of infants who would sit upon treatment was 20% and, using an exact binomial test with a 1-sided 5% significance level, that a target sample size of 40 infants would provide at least 90% power to test the null hypothesis H₀:P of 5% or less versus the alternative hypothesis H_a:P of more than 5%. As such, a minimum of 5 infants sitting without support would provide a statistically significant difference from the predefined performance criterion. Which would translate into the lower limit of the 2-sided 90% Clopper-Pearson (exact) CI would be above 5%.

To adjust for multiple testing, the sponsor implemented a hierarchical testing approach where hypothesis testing in secondary outcomes would be implemented only if the primary outcome was significant (the proportion of infants who are sitting without support after 12 months of treatment). Then, testing (at 1-sided 5% significance) would adhere to the following hierarchy, where testing cannot proceed beyond a non-statistically significant outcome.

• Proportion of patients who achieve a score of 40 or higher in the CHOP INTEND at month 12: This will be contrasted with a predefined threshold of 17%, which was informed by the Network for Excellence in Neuroscience Clinical Trials (NeuroNEXT) SMA infant biomarker study.⁴²

• Proportion of patients who achieve an increase of at least 4 points in their CHOP INTEND score from baseline at month 12: This will be contrasted with a predefined threshold of 17%, which was informed by the NeuroNEXT SMA biomarker study.⁴²

• Proportion of motor milestone responders as assessed by the HINE Section 2 at month 12: This was defined as an improvement in a motor milestone of at least a 2-point increase in ability to kick (or maximal score) or a 1-point increase in head control, rolling, sitting, crawling, standing, or walking. This will be contrasted with a predefined threshold of 12%, which was based on the results from a study conducted by De Sanctis et al. (2016).⁴

• Proportion of patients alive without permanent ventilation at month 12: This will be contrasted with a predefined threshold of 42%, which was informed by the NeuroNEXT SMA biomarker study.⁴²

• Proportion of patients sitting without support for 30 seconds (item 26 of BSID-III) at month 24: This will be contrasted with a predefined threshold of 0%, which was based on the results from a study conducted by De Sanctis et al. (2016).⁴

• Proportion of patients standing (item 40 of BSID-III) at month 24: This will be contrasted with a predefined threshold of 0%, which was based on the results from a study conducted by De Sanctis et al. (2016).⁴

• Proportion of patients walking (item 42 of BSID-III) at month 24: This will be contrasted with a predefined threshold of 0%, which was based on the results from a study conducted by De Sanctis et al. (2016).⁴

In a similar manner to the primary outcome, the threshold for the null hypothesis was based on the results in natural history studies of a similar population. The sponsor outlined that these tests, at a 1-sided 5% significance level, would only be performed if a predefined benchmark is determined from historical data.

These thresholds were established based on a literature review that the sponsor conducted to identify data describing the natural history of infants with type I SMA. The sponsor aimed to derive these minimum thresholds by selecting a population of type I SMA in the literature that is similar to the 1 included in FIREFISH Part 2. To that aim, the following factors were considered when determining the extent of similarity: 1) the SMN2 copy number, 2) age at symptom onset, 3) age at enrolment, 4) type I classification, 5) similarities in the standard of care, 6) the time period, 7) the region, 8) the type of centre. Through an unclear operationalization mechanism, the sponsor has decided that the NeuroNEXT SMA infant biomarker study was the study most similar to the infants enrolled in FIREFISH Part 2.⁴ In cases where the outcome was not described in the NeuroNEXT study, the sponsor used the study by De Sanctis et al. (2016) to inform on the minimum threshold. The determination of the threshold was based on the upper limit of the 90% CI in the outcomes of NeuroNEXT or De Sanctis et al. (2016).

In categorical outcomes, infants who do not achieve the outcome, do not maintain a previously achieved outcome at the time of assessment, have been withdrawn from the study, died, or have a missing assessment at the time of the outcome are all considered as not having achieved the outcome. The proportion in categorical outcomes was presented with a 2-sided 90% Clopper-Pearson (exact) CI.

Other continuous measures (e.g., change in CHOP INTEND score from baseline) would use observed data only for the main analysis, with no imputation of missing data.

• The median time to event and the proportion of infants without an event at a certain time point was to be estimated using the Kaplan-Meier methodology. Infants who had been withdrawn from the study with no event reported before withdrawal were to be censored at the date of withdrawal. For infants who had been withdrawn from the study and entered into safety follow-up, all events reported from the date of enrolment up to the date of withdrawal were to be included in the analysis.

Several sensitivity analyses were outlined in the protocol for FIREFISH outcomes. These included the following.

1. Site evaluation of sitting without support for 5 seconds: The outcomes will be based on the assessment of the clinical site evaluator instead of the assessment of 2 independent central readers.

2. Alternative definition of sitting without support: Sitting will be defined by the HINE Section 2 categories of “stable sit” or “pivots (rotates).” Infants within either of these response categories for the milestone of sitting at month 12 will be classified as responders.

3. Assuming patients who have been withdrawn from the study without entering safety follow-up as events in the time to death or permanent ventilation analysis.

4. In patients who have been withdrawn but were entered in the safety follow-up, including any observed event during the safety follow-up in the time to death or permanent ventilation analysis.

5. Applying the following method for handling missing data in CHOP INTEND–related outcomes: Missing data at baseline will be imputed with the median of an infant’s population stratum (symptom onset ≤ 3 months, or > 3 months), and missing month 12 data will be imputed with a linear interpolation of the values of neighbouring visits, if available, or with the minimum value of the population stratum that the infant belongs to, if neighbouring values are not available.

Subgroup analyses were planned for the following outcomes: the proportion of infants sitting without support for 5 seconds at month 12, the time to death or permanent ventilation, and the proportion of infants who achieved a CHOP INTEND score of 40 or higher at month 12. It included the following subgroups that were specific to our protocol: baseline CHOP INTEND score (≤ median score and > median score), and time between first treatment and the onset of symptoms (≤ 3 months and > 3 months).

SUNFISH Part 2

SUNFISH Part 2 was a randomized, double-blind, placebo-controlled trial. The sample size for SUNFISH Part 2 was calculated with the assumption that the true treatment difference of risdiplam versus placebo would be 3 points in the total MFM-32 score with a common SD of 6. Based on this, a sample size of 168 patients, with a 10% dropout rate allowance, would provide at least 80% power at a 2-sided 5% significance level for testing the null hypothesis (the true treatment difference is 0).

The randomization was stratified based on age groups. These were 2 years to 5 years, 6 years to 11 years, 12 years to 17 years, and 18 years to 25 years at randomization. The sponsor aimed to randomize a maximum of 30 patients into the 18 years to 25 years age group, while a minimum of 45 patients would be randomized into each of the other 3 age groups. No clear justification was provided for this stratification approach.

The sponsor indicated that, in addition to the primary outcome, hypothesis testing was performed on 6 key secondary outcomes. To control for type I error, a gatekeeping approach was applied to the 7 null hypotheses that were grouped into 6 families. Testing was conducted on a hierarchical order with a truncated Hochberg procedure used within families with more than a single hypothesis. An unadjusted P value was the P value obtained when an end point was tested at a 2-sided 0.05 significance level. The adjusted P values were derived based on all the (unadjusted) P values from end points in order of the hierarchy up to the current end point. The hierarchy of and the outcomes in the families were as follows.

1. Family 1 includes the hypothesis for the primary end point on the change from baseline total MFM-32 score at month 12 comparing risdiplam with placebo: H₁₁ (MFM-32).

2. Family 2 includes the hypothesis for the proportion of patients who achieve a change from baseline of 3 or more on the total MFM-32 score at month 12 comparing risdiplam with placebo: H₂₁ (proportion with an MFM-32 ≥ 3).

3. Family 3 includes the hypothesis for the change from baseline total score of RULM at month 12 comparing risdiplam with placebo: H₃₁ (RULM).

4. Family 4 includes the hypothesis for the change from baseline total score of HFMSE at month 12 comparing risdiplam with placebo and also the hypothesis for the change from baseline best percentage predicted value in FVC at month 12 comparing risdiplam with placebo: H₄₁ (HFMSE), H₄₂ (FVC).

5. Family 5 includes the hypothesis for the change from baseline total score of SMAIS at month 12 comparing risdiplam with placebo: H₅₁ (SMAIS).

6. Family 6 includes the hypothesis for the proportion of patients rated by the clinician as improved in the CGI-C scale at month 12 comparing risdiplam with placebo: H₆₁ (CGI-C).

For the primary outcome as well as other outcomes that measure change from baseline, the sponsor used a mixed model of repeated measures analysis that contains fixed effects, random effects, and random error components. In the mixed model of repeated measures model, the dependent variable was the absolute change from baseline score. This was tested at a 2-sided 5% level of significance. The fixed effects variables were baseline score, treatment group, time, treatment-by-time interaction, baseline-by-time interaction, and the randomization stratification age group. Random effect included the patient effect and time as a repeated variable within a patient. Patient, treatment, and time were treated as factor variables and the baseline total MFM-32 score as a covariate. Within-patient variability was modelled with an unstructured variance-covariance matrix. The components of the matrix were estimated by the restricted maximum likelihood method and the degrees of freedom estimated using the Kenward-Roger approximation.

For categorical outcomes representing responder analysis, a logistic regression model was used for the analysis of these end points. The logistic regression model included the baseline total score, treatment, and age group.

Missing data of items within the Motor Function Measure (MFM) were handled through the following rule: D1, D2, and D3 were only calculated if there was less than 15% of missing data, and total scores were only calculated if scores were available for each domain. Missing items that constitute less than 15% of the domain were imputed as 0. Missing MFM total scores were not imputed. For the RULM, if 3 or fewer items are missing, they were imputed as zero, while the RULM total score was not calculated if there were more than 3 missing items. For the HFMSE, if 6 or fewer items are missing, they were imputed as zero, while the HFMSE total score was not calculated if there were more than 6 missing items.

A sensitivity analysis was planned if more than 5% of the intention-to-treat (ITT) population was excluded from the MFM-32 primary end point analysis due to missing data. The sensitivity analysis imputed missing MFM as follows: missing items at baseline were imputed with values from the screening visit, missing items at week 17 or week 35 were imputed by the mean value of the 2 neighbouring visits, and missing items at week 52 were imputed with values from the previous visit. If there were still missing data following this approach, the remaining missing data were to be treated as per the rule for the base MFM-32 analysis.

A treatment policy strategy analysis was also planned to include all data for patients who discontinued the treatment but remained in the study, regardless of whether or not they had initiated a prohibited therapy.

Additional sensitivity analysis to assess the impact of missing data on the primary end point due to study withdrawals was planned. The sensitivity analysis was a tipping point analysis where patients who had monotone missing data were assumed to have worse efficacy outcomes by some amount of delta compared to similar patients with observed data. The analysis was to be performed on a series of increasing values of delta until the analysis conclusion of statistically significant finding no longer held.

Subgroup analysis was performed over the primary outcome and the key secondary outcomes (the secondary outcomes included in the statistical testing hierarchy). It included the following subgroups that were pre-specified in the protocol for this review: age groups (2 years to 5 years, 6 years to 11 years, 12 years to 17 years, 18 years to 25 years at randomization), disease severity (MFM-32 baseline total score first quartile, second and third quartile, and fourth quartile), SMA type (type II and type III), and SMN2 copy number (< 2, 2, 3, ≥ 4, unknown).

A summary of the statistical analysis approach of the efficacy end points in FIREFISH Part 2 and SUNFISH Part 2 is provided in Table 10.

Analysis Populations

FIREFISH Part 2

• ITT: Defined as all infants enrolled in FIREFISH Part 2, regardless of whether they received the treatment or not. This population is used in all efficacy analysis.

• Safety population: Defined as all infants who received at least 1 dose of risdiplam.

SUNFISH Part 2

• ITT: Defined as all patients randomized in SUNFISH Part 2. This population is used in all efficacy analysis.

• Safety population: Defined as all patients who received at least 1 dose of risdiplam or placebo.

Results

Patient Disposition

FIREFISH Part 2

Table 11 summarizes the disposition of enrolled infants. A total of 52 infants were screened; 11 were determined to be a screen failure and 41 patients were administered risdiplam. Of the 11 screen failures, 4 infants died during the screening period and 3 did not meet all inclusion criteria (these 2 categories were the most common reasons for the screen failures). Of the 41 enrolled patients, 38 (92.7%) had completed 12 months of treatment and none have yet completed 24 months of treatment.

SUNFISH Part 2

Table 11 summarizes the disposition of enrolled patients in SUNFISH. A total of || patients were screened; 31 patients were determined to be a screen failure and 180 patients were randomized to risdiplam (n = 120) and placebo (n = 60). Most common reasons for the screen failures were that 5 patients did not achieve the required baseline MFM-32 scores, 5 patients had reasons related to inability to measure their optical coherence tomography, 4 patients were diagnosed with febrile conditions that excluded them from the study, and 3 patients withdrew their consent. Of 120 patients who were randomized into the risdiplam arm, 3 withdrew. Of the 60 patients randomized to the placebo arm, 1 withdrew. The most common reason for withdrawal was access to other SMA treatment.

Exposure to Study Treatments

FIREFISH Part 2

All enrolled patients received risdiplam at a median dose intensity (the number of doses actually received divided by the expected number of doses) of 100% (range 98.0% to 100%). As of the clinical cut-off date of the included FIREFISH Part 2 report, of the ITT population (including the patients who died or withdrew) the median treatment period was 15.2 months (range = 1.6 months to 10.2 months).

Most patients received some form of concomitant medication. This included |||| who received medications for AEs, |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ||||||||||||||||||||||||||||||||||||||

SUNFISH Part 2

The median risdiplam dose intensity was |||||||||||||||||||||||||||||||||||||||||| and |||||| of patients had a dose intensity ||||. As of the clinical cut-off date of the included SUNFISH Part 2 report, the median treatment period was |||||||||||||||||||||||||||||||||||||||| ||||.

|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||

Efficacy: FIREFISH Part 2

Only those efficacy outcomes and analyses of subgroups identified in the review protocol are reported as follows. Table 12 provides a summary of the efficacy results of FIREFISH Part 2.

Primary Outcome: Proportion of Infants Sitting Without Support for 5 Seconds (Item 22 of BSID-III) at Month 12

At 12 months of treatment with risdiplam, 29.3% of infants (12 out of 41 infants; 90% CI, 17.8 to 43.1) were able to able to sit without support for 5 seconds. This proportion was statistically significantly higher than the predefined performance criterion of 5% based on natural history data from the NeuroNEXT cohort (P < 0.0001).

Sensitivity analysis using the assessment of the clinical site indicated that 34.1% of infants (14 out of 41 infants; 90% CI, 22.0 to 48.1) achieved sitting without support for 5 seconds, while a sensitivity analysis using the HINE Section 2 categories of “stable sit” and “pivots (rotates)” indicated that 24.4% of the infants (10 out of 41 infants; 90% CI not available) achieved the outcome.

Subgroup analysis showed that of 14 infants with a disease duration of 3 months or less, |||||||||||||||||||||||||||||||||||||||||||||||||| achieved the primary outcome, while of 27 infants with a disease duration of more than 3 months, |||||||||||||||||||||||||||||||||||||||| |||| achieved the primary outcome. Subgroup analysis based on the baseline CHOP INTEND score showed that 28.6% of the infants (6 out of 21 infants; 90% CI, 13.24 to 48.74) with a score equal to the median or below (≤ 22) achieved the primary outcome, and 30.0% of the infants (6 out of 20 infants; 90% CI, 13.96 to 50.78) with a CHOP INTEND score over the median (> 22) achieved the primary outcome.

Motor Function–Related Outcomes

Since none of the patients have reached the 24 months of treatment end point, only the results at 12 months of treatment are reported.

Of the reported secondary outcomes that are within the statistical hierarchy (at 12 months of treatment), 56.1% of infants (23 out of 41 infants; 90% CI, 42.1 to 69.38) had a CHOP INTEND score of 40 or higher (P < 0.0001 against a performance criterion of 17%), 90.2% of infants (37 out of 41 infants; 90% CI, 79.1 to 96.6) have achieved an increase of at least 4 points in the CHOP INTEND score from baseline (P < 0.0001 against a performance criterion of 17%), and 78.0% of infants (32 out of 41 infants; 90% CI, 64.8 to 88.0) were considered motor milestone responders assessed through the HINE Section 2 (P < 0.0001 against a performance criterion of 12%).

For CHOP INTEND–related outcomes, sensitivity analysis of using a data imputation approach based on the median of the infant’s stratum and linear interpolation of flanking values showed results that did not differ from the base case.

Subgroup analysis showed that of 14 infants with a disease duration of 3 months or less, |||||||||||||||||||||||||||||||||||||||||||||||||||| achieved a CHOP INTEND score of 40 or higher, while of 27 infants with a disease duration of more than 3 months, |||||||||| |||||| |||||||||||||||||||||||||| achieved a CHOP INTEND score of 40 or higher. Subgroup analysis based on baseline CHOP INTEND score showed that 42.9% of the infants (9 out of 21 infants; 90% CI, 24.50 to 62.81) with a score equal to the median or below (≤ 22) achieved a CHOP INTEND score of 40 or higher, and 70.0% of the infants (14 out of 20 infants; 90% CI, 49.22 to 86.04) with a CHOP INTEND score over the median (> 22) achieved a CHOP INTEND score of 40 or higher.

Other motor function–related outcomes, at month 12 treatment, were outside the statistical hierarchy. Results demonstrated that 53.7% of infants achieved head control as assessed through CHOP INTEND item 12, 22.0% were able to support weight or stand with support as assessed by the HINE Section 2, 2.4% were able to bounce while assessing the walking item of the HINE Section 2, |||| were able to sit without support for 30 seconds as assessed by item 26 in BSID-III, and patients showed a mean change of 7.21 (SD = 5.71) points in the raw BSID-III overall score from baseline.

Survival-Related Outcomes

The proportion of infants alive without permanent ventilation at month 12 was the secondary outcome within the statistical testing hierarchy. A total of 85.4% of patients (90% CI, 73.4 to 92.2) were alive and did not require permanent ventilation at month 12 (35 out of 41 patients). This outcome was contrasted with a predefined threshold of 42% to give a statistically significant finding with a P value of less than 0.0001.

Since all study withdrawals were due to an event, planned sensitivity analysis of the impact of a non-event withdrawal of an infant was not presented.

Subgroup analysis showed that of 14 infants with a disease duration of 3 months of less, |||||||||||||||||||||||||||||||||| were event free (death or permanent ventilation) by month 12, while of 27 infants with a disease duration of more than 3 months, ||||||||||||||||||||||||||||| event free by month 12. Subgroup analysis based on baseline CHOP INTEND showed ||||||||||||||||||||||||||| of the infants with a CHOP INTEND score over the median (> 22) were event free by month 12.

Other survival-related outcomes were outside the statistical testing hierarchy. The proportion of patients alive at month 12 was 92.7% (38 out of 41 patients). Median time to event was not presented due to there being too few events to allow its estimation.

Respiratory-Related Outcomes

Outcomes were outside the statistical testing hierarchy, meaning that the type I error rate was not controlled for these statistical tests. The proportion of infants without permanent ventilation at month 12 of treatment was 92.7% (median time to permanent ventilation was not possible to estimate due to few available events). The proportion of infants at month 12 who did not require any type of ventilatory support was 24.4% (10 out of 41 infants).

Nutrition-Related Outcomes

At 12 months of treatment with risdiplam, 82.9% of the enrolled infants (34 out of 41 infants) had the ability to feed orally. This outcome was outside the statistical testing hierarchy.

Health Care Utilization–Related Outcomes

At 12 months of treatment with risdiplam, 48.8% of the enrolled infants (20 out of 41 infants) did not require hospitalization. Hospitalization was defined as any admission that lasted for more than 2 days. Hospitalizations per patient-year were estimated at 1.30 (90% CI, 1.02 to 1.65). These outcomes were outside the statistical testing hierarchy.

Health-Related Quality of Life–Related Outcomes

From baseline to month 12 of treatment, there was a mean change of |||||||||||||||||| in the ITQOL-SF47 overall health domain score, a mean change of |||||||||||||||||| in the ITQOL-SF47 parental impact (emotional) domain score, and a mean change of |||||||||||||| in the ITQOL-SF47 parental impact (time) domain score.

Efficacy: SUNFISH Part 2

Only those efficacy outcomes and analyses of subgroups identified in the review protocol are reported as follows. Table 13 provides a summary of the efficacy results of SUNFISH Part 2

Primary Outcome: Change From Baseline in MFM-32 Total Score at Month 12

Valid baseline MFM-32 baseline values were available for 115 patients and 59 patients in the risdiplam and placebo arms, respectively. Patients receiving risdiplam had a mean change from baseline at month 12 of 1.36 points (95% CI, 0.61 to 2.11) as opposed to a mean change in the placebo group of –0.19 (95% CI, –1.22 to 0.84). This translates into a mean difference from placebo of 1.55 points (95% CI, 0.30 to 2.81; P = 0.0156).

|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||

Subgroup analyses, presented in Figure 4, suggest that the only groups where the treatment effect excluded the null was observed in the youngest age group (2 years to < 6 years) and in the group with the least disease severity at baseline (MFM-32 total score at baseline ≤ first quartile [Q1]). Overall, all estimates show wide CIs. Numerically, the point estimate of the treatment effect decreased, and the range of the CIs increased as we moved into older age groups.

Figure 4: Subgroup Analyses of Change From Baseline in MFM-32 Total Score at Month 12

CI = confidence interval; MFM-32 = Motor Function Measure–32 items; SMA = spinal muscular atrophy; SMN2 = survival of motor neuron 2.

Source: Drug Reimbursement Review sponsor submission.¹⁴

Motor Function–Related Outcomes

Since none of the patients have reached the 24 months of treatment end point, only the results at 12 months of treatment are reported.

Change From Baseline of 3 or More in MFM-32 Total Score at Month 12 (Family 2)

This outcome was the second end point to be tested within the statistical testing hierarchy. Under this definition, 38.3% of patients in the risdiplam arm (44 out of 115 patients) were considered responders, while 23.7% of patients in the placebo group (14 out of 59 patients) were considered responders. The outcome was statistically significant with an OR of 2.35 (95% CI, 1.01 to 5.44) for risdiplam versus placebo.

Figure 5: Subgroup Analyses of Change From Baseline of 3 or More in MFM-32 Total Score at Month 12 (Family 2)

Figure 5 was removed upon request from the sponsor because it contained confidential information.

MFM-32 = Motor Function Measure–32 items.

Source: Drug Reimbursement Review sponsor submission.¹⁴

Change From Baseline of 0 or More in MFM-32 Total Score at Month 12

This outcome was outside the statistical testing hierarchy, meaning that no control for type I error was performed. A total of 69.6% of patients in the risdiplam arm (80 out of 115 patients) achieved this end point, while a total of 54.2% of patients in the placebo group (32 out of 59 patients) achieved it. The difference translates into an OR of 2.00 (95% CI, 1.02 to 3.93) for risdiplam versus placebo.

Change From Baseline in MFM-32 Domain 1 Score at Month 12

|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| 　|　 ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||

Change From Baseline in MFM-32 Domain 2 Score at Month 12

|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| 　|　 ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||

Change From Baseline in MFM-32 Domain 3 Score at Month 12

|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| 　|　 ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||

Change From Baseline in MFM-32 Domain 1 and Domain 2 Scores at Month 12

|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| 　|　 ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||

Change From Baseline in MFM-32 Domain 2 and Domain 3 Scores at Month 12

|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| 　|　 ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||

Change From Baseline in RULM Total Score at Month 12 (Family 3)

This outcome was the third end point to be tested within the statistical testing hierarchy. Valid RULM baseline values were available for 119 patients and 58 patients in the risdiplam and placebo arms, respectively. Patients receiving risdiplam had a mean change from baseline at month 12 of 1.61 points (95% CI, 1.00 to 2.22) as opposed to a mean change in the placebo group of 0.02 (95% CI, –0.83 to 0.87). The mean difference from placebo was 1.59 points (95% CI, 0.55 to 2.62; P = 0.0028).

Subgroup analyses, presented in Figure 6, show wide CIs in all the estimates and indicate similar patterns related to age groups that were observed in the primary outcome.

Figure 6: Subgroup Analyses of Change From Baseline in RULM Total Score at Month 12

CI = confidence interval; RULM = Revised Upper Limb Module; SMA = spinal muscular atrophy; SMN2 = survival of motor neuron 2.

Source: Drug Reimbursement Review sponsor submission.¹⁴

Change From Baseline of 0 or More in RULM Total Score at Month 12

This outcome was outside the statistical testing hierarchy. |||| of patients in the risdiplam arm |||||||| achieved this end point, while |||| of patients in the placebo group |||||| achieved it. The difference translates into an OR of |||||||||||||| |||||||||||| for risdiplam versus placebo.

Change From Baseline of 2 or More in RULM Total Score at Month 12

This outcome was outside the statistical testing hierarchy. A total of 47.9% of patients in the risdiplam arm (57 out of 119 patients) achieved this end point, while 31.0% of patients in the placebo group (18 out of 58 patients) achieved it. The difference translates into an OR of 2.18 (95% CI, 1.05 to 4.54) for risdiplam versus placebo.

Change From Baseline in HFMSE Total Score at Month 12 (Family 4)

This outcome was 1 of 2 end points within the fourth family of outcomes to be tested in the statistical hierarchy. Valid HFMSE baseline values were available for all patients. Patients receiving risdiplam had a mean change from baseline at month 12 of 0.95 (95% CI, 0.29 to 1.61) as opposed to a mean change in the placebo group of 0.37 (95% CI, –0.54 to 1.28). Mean difference from placebo was 0.58 points (95% CI, –0.53 to 1.69), which was statistically nonsignificant. As such, all subsequent end points in the statistical hierarchy should no longer be tested against a null hypothesis.

Figure 7: Subgroup Analyses of Change From Baseline in Total Score of the HFMSE at Month 12

Figure 7 was removed upon request from the sponsor because it contained confidential information.

HFMSE = Hammersmith Functional Motor Scale Expanded.

Source: Drug Reimbursement Review sponsor submission.¹⁴

Change From Baseline of 0 or More in HFMSE Total Score at Month 12

This outcome was outside the statistical testing hierarchy. |||| of patients in the risdiplam arm |||||||| achieved this end point, while |||| of patients in the placebo group |||||| achieved it. The difference translates into an OR of |||||||||||||| |||||||||||| for risdiplam versus placebo.

Change From Baseline of 2 or More in HFMSE Total Score at Month 12

This outcome was outside the statistical testing hierarchy. |||| of patients in the risdiplam arm |||||||| achieved this end point, while |||| of patients in the placebo group |||||| achieved it. The difference translates into an OR of |||||||||||||| |||||||||||| for risdiplam versus placebo.

Survival-Related Outcomes

No survival-related outcome was reported in the SUNFISH Part 2 study.

Respiratory-Related Outcomes

Change From Baseline in Best Percentage Predicted Value of Forced Vital Capacity at Month 12 for Family 4

Spirometry was performed on patients 6 years of age or older at the time of screening (83 in the risdiplam arm and 42 in the placebo arm). The FVC outcome was 1 of 2 end points within the fourth family of outcomes to be tested in the statistical hierarchy. Valid FVC baseline values were available for 83 patients and 40 patients in the risdiplam group and placebo group, respectively. Patients receiving risdiplam had a mean change from baseline at month 12 of –5.16 (95% CI, –7.93 to –2.39) as opposed to a mean change in the placebo group of –3.11 (95% CI, –6.95 to 0.74). The mean difference from placebo was –2.05 (95% CI, –6.67 to 2.56), which was statistically nonsignificant.

Figure 8: Subgroup Analyses of Change From Baseline in Best Percentage Predicted Value of FEV₁ at Month 12

Figure 8 was removed upon request from the sponsor because it contained confidential information.

FEV₁ = forced expiratory volume in 1 second.

Source: Drug Reimbursement Review sponsor submission.¹⁴

Change From Baseline in Best Percentage Predicted Value of Forced Expiratory Volume in 1 Second at Month 12

This outcome was outside the statistical testing hierarchy. Patients receiving risdiplam had a mean change from baseline at month 12 of –4.22 (95% CI, –7.49 to –0.96) as opposed to a mean change in the placebo group of –1.35 (95% CI, –5.91 to 3.20). This translates into a mean difference of –2.87 (95% CI, –8.36 to 2.62).

Nutrition-Related Outcomes

No nutrition-related outcome was reported in the SUNFISH Part 2 study.

Health Care Utilization–Related Outcomes

No related outcome was reported in the SUNFISH Part 2 study.

Health-Related Quality of Life–Related Outcomes

Change From Baseline in Caregiver-Reported SMAIS Total Score at Month 12 (Family 5)

This outcome was the fifth end point to be tested within the statistical testing hierarchy. However, since the fourth family of outcomes in the statistical hierarchy failed to achieve statistical significance, hypothesis testing was no longer valid for this, or subsequent outcomes, and any reported P value was nominal in nature and type I error was no longer controlled for in such outcomes. Valid baseline values were available for 116 patients and 60 patients in the risdiplam group and placebo group, respectively. Patients receiving risdiplam had a mean change from baseline at month 12 of 1.65 (95% CI, 0.66 to 2.63) as opposed to a mean change in the placebo group of –0.91 (95% CI, –2.23 to 0.42). The mean difference from placebo was 2.55 (95% CI, 0.93 to 4.17).

Change From Baseline in Patient-Reported SMAIS Total Score at Month 12

Valid baseline values were available for 43 patients and 23 patients in the risdiplam and placebo group, respectively. Patients receiving risdiplam had a mean change from baseline at month 12 of 1.04 (95% CI, –0.26 to 2.35) as opposed to a mean change in the placebo group of –0.40 (95% CI, –2.13 to 1.32). The mean difference from placebo was 1.45 (95% CI, –0.68 to 3.57).

Proportion of Individuals Rated as “Improved” on the CGI-C Scale at Month 12 (Family 6)

This outcome was the sixth and last end point to be tested within the statistical testing hierarchy. However, since the fourth family of outcomes in the statistical hierarchy failed to achieve statistical significance, hypothesis testing is no longer valid for this outcome, as there is no control for type I error. Valid baseline values were available for all patients. In all, 47.5% of patients in the risdiplam arm (57 out of 120 patients) achieved this end point, while 40.0% of patients in the placebo group (24 out of 60 patients) achieved it. The difference translates into an OR of 1.38 (95% CI, 0.70, 2.74) for risdiplam versus placebo.

Harms: FIREFISH Part 2

Only those harms identified in the review protocol are reported as follows. See Table 14 for detailed harms data.

Adverse Events

At least 1 AE was reported in all enrolled infants. Upper respiratory tract infection was the most commonly reported AE (46.3%), followed by pneumonia (39.0%), pyrexia (39.0%), and constipation (19.5%).

Serious Adverse Events

SAEs were reported in 58.5% of the infants (24 out of 41 infants). The majority of the SAEs were related to respiratory problems or respiratory infections.

Withdrawals Due to Adverse Events

There were no withdrawals due to AEs (WDAEs).

Mortality

Three infants died during the study. Two deaths were attributed to pneumonia, and 1 to respiratory failure.

Notable Harms

Early animal studies with risdiplam suggested potential ophthalmic complications associated with treatment with risdiplam. Ophthalmic assessment in FIREFISH Part 2 identified 1 ophthalmic AE (heterophoria).

Harms: SUNFISH Part 2

Only those harms identified in the review protocol are reported as follows. See Table 14 for detailed harms data.

Adverse Events

At least 1 AE was reported in 92.5% and 91.7% of enrolled patients in the risdiplam and placebo arms, respectively. Upper respiratory tract infection was the most commonly reported AE (31.7% in the risdiplam arm and 30.0% in the placebo arm), followed by nasopharyngitis (25.8% in the risdiplam arm and 25.0% in the placebo arm), pyrexia (20.8% in the risdiplam arm and 16.7% in the placebo arm), and headache (20.0% in the risdiplam arm and 16.7% in the placebo arm).

Serious Adverse Events

SAEs were reported in 20.0% of patients who received risdiplam and 18.3% in patients who received placebo. Most of the SAEs were related to respiratory problems or respiratory infections.

Withdrawals Due to Adverse Events

There were no WDAEs.

Mortality

There were no deaths reported.

Notable Harms

Early animal studies with risdiplam suggested potential ophthalmic complications associated with treatment with risdiplam. Ophthalmic assessment in SUNFISH Part 2 did not show any findings suggestive of risdiplam-induced effects.

Critical Appraisal: FIREFISH Part 2

Internal Validity

FIREFISH Part 2 was a single-arm, open-label, uncontrolled, phase III clinical trial that aimed to assess the efficacy and safety of risdiplam in infants with type I SMA. Specifically, patients were aged 1 month to 7 months with homozygous or heterozygous deletion or point mutation in the SMN1 gene, showing signs and symptoms attributed to type I SMA, and had 2 SMN2 gene copies. The study used information from the literature on the natural history of type I SMA to establish a minimum threshold against which FIREFISH Part 2 outcomes were tested. The study reported minimal missing data. Furthermore, a hierarchal statistical testing model was implemented to adjust for multiple testing in some secondary outcomes.

Limitations to the internal validity of FIREFISH Part 2 revolve around the following points:

• Lack of a control group

  • In the absence of a concurrent control arm in the form of a placebo control or an active control, there is a potential for bias and confounding (i.e., an overestimate) in the treatment effect estimated with risdiplam in the single-arm trial FIREFISH Part 2. Without a randomized comparison to a control group, natural fluctuations in the disease cannot be controlled for, nor can the effects of known or unknown confounders. The sponsor justified the study design on the strong unmet need in the population, the poor prognosis of the disease, and the objective nature of the primary outcome. It should be noted that the sponsor method of determining a minimum threshold for testing the null hypothesis does not reduce the risk of bias and confounding in the same manner as randomization to a concurrent control arm. While basing these thresholds on well-described literature may be useful to reflect the natural history of the disease, this approach has many limitations that arise due to variations in study characteristics, measurement methods, patients’ characteristics, standards of care, and other potentially unknown or unmeasured confounders. The sponsor derived these thresholds from the upper 90% CI in the selected natural history studies. No other method of adjusting for potential differences between the single-arm trial participants and historical control was attempted. These limitations involving the study design and the method of establishing the null hypothesis threshold led to a level of uncertainty regarding the extent of the magnitude of the observed clinical effect that is solely attributed to risdiplam.

• Lack of masking for intervention or outcome assessment

  • Investigators, outcome assessors, and parents were aware that infants received risdiplam. This may increase the risk of observer and respondent bias (on the part of the proxy) in assessing outcome measures. The potential risk of observer bias may not be strong in cases of objective outcomes, such as death and permanent ventilation. The sponsor video-recorded the assessment of motor milestone outcomes and used 2 independent central reviewers to determine the outcomes. The lack of blinding remains a concern for outcomes beyond death or permanent ventilation, especially in patient-reported (on the part of the proxy) outcomes.

• Lack of imputation of missing data in continuous outcomes

  • While missing data were treated as a nonresponder in categorical outcomes, no method for imputing missing data were employed in continuous outcomes. These outcomes included a change in BSID-III score from baseline (||||||||||), and the ITQOL-SF47 results (||||||||||||). The sponsor conducted a sensitivity analysis to assess the potential impact of missing data on the CHOP INTEND–related outcomes, but no similar approach was described for the rest of the outcomes. While the sensitivity analysis of CHOP INTEND did not indicate that missing data have affected the outcome, it is unclear if this is also the case in the change in BSID-III score and in the ITQOL-SF47 outcomes.

• Modified use of motor function assessment tools

  • The sponsor used the BSID-III in a modified way that was not part of the original development of the tool. Although the sponsor indicated that these modifications were established with the help of the original tool developers, the extent at which the validity assessment of the original tool applies to the modified tool is unclear. Similarly, the HINE Section 2 was used with the exclusion of the voluntary grasp item and with a definition of motor responders that is not assessed for validity on the original tool. Additionally, there is no established MID regarding the BSID-III raw score or the quantifiable change in HINE Section 2 scores that is the basis of the definition of motor responder.

• The use of a 1-sided 5% significance testing

  • The sponsor used a 1-sided 5% significance level for determining the statistical significance of the primary outcome. This approach provides a larger threshold to accept treatment differences as causative when compared to a 2-sided test or a 1-sided test at a 2.5% significance level. External Validity

The inclusion and exclusion criteria of FIREFISH Part 2 enrolled patients who are, or most likely to be, diagnosed with type I SMA and are at an early stage of the disease. The outcomes used to inform on the efficacy are clinically meaningful and reflective of clinical practice. They are, specifically, sitting without support, death or permanent ventilation, and other motor functions and respiratory measures.

Limitations to the external validity of FIREFISH Part 2 revolve around the following points:

• Narrow inclusion criteria

  • FIREFISH Part 2 did not include infants who were pre-symptomatic, who were 7 months of age or older, or who had more than 2 copies of SMN2. Other SMA infants who were not included in the study may have had a more severe presentation of the disease (e.g., older SMA type I patients, patients requiring ventilatory support) and the results of FIREFISH Part 2 cannot inform on the extent of potential benefit offered by risdiplam for these patients. In addition, there were no infants younger than 2 months enrolled in the study.

• Changes in the clinical practice and the lack of a direct head-to-head comparison to other disease-modifying agents

  • The clinical trial program investigating risdiplam had started when other SMA disease-modifying agents were not yet available for patients with SMA. At the moment, in Canada, patients with a diagnosed SMA who meet the eligibility criteria for treatment with nusinersen may have already received several doses of nusinersen. In addition, onasemnogene abeparvovec has recently gained market authorization in Canada. As such, the thresholds for efficacy that were used to inform the null hypothesis of various outcomes is no longer truly reflective of the disease progression in the current clinical practice in Canada, and currently treated infants are now expected to live longer and gain certain motor milestones. The results of the efficacy of risdiplam, presented in FIREFISH Part 2, cannot inform on its potential comparative efficacy to nusinersen or onasemnogene abeparvovec.

• The use of sitting unsupported for 5 seconds as a primary outcome versus 30 seconds

  • According to the experts consulted for this review, the 30 seconds is a more clinically relevant outcome than sitting unsupported for 5 seconds. As such, it is not clear why the 5 seconds was chosen over the 30 seconds. There is also a concern that the assessment of sitting unsupported for 5 seconds is less objective than 30 seconds. Although the sponsor has implemented an independent review process in the assessment of the primary outcome, the lack of a control arm limits our ability to judge possible biases in the assessment procedure.

• Lack of Canadian sites

  • No Canadian sites participated in FIREFISH Part 2. While there is no clear reason to assume major differences in the standards of practice and the patient population between the US and Canada, the lack of representation of Canadian patients could potentially reduce the generalizability to Canadian settings if practice variation exists.

• Assessment period of 12 months

  • Considering the lifelong nature of SMA, and the narrow SMA population in FIREFISH Part 2, generalizability of the efficacy and safety results beyond the presented 12 months remains uncertain. It is, however, understood that the study is ongoing until it has reached the 24-month mark and will continue into an open-label extension study until the time of drug commercialization.

• Lack of a clearly established MID in the total scores of the motor function tools used in FIREFISH Part 2

  • The sponsor provided several motor function–related outcomes that are based on achieving a certain threshold of total points or points difference versus baseline in the CHOP INTEND, HINE Section 2, and BSID-III. With the possible exception of HINE Section 2, the CHOP INTEND and BSID-III do not have a clearly established MID for total scores or score difference. This limits our ability to understand the extent of the clinical significance of the observed effect.

• The use of motor function tools that are not commonly conducted in clinical practice

  • The clinical experts consulted on this review indicated that the assessment of motor function may vary from centre to centre, with no uniform approach to motor function assessment throughout Canada. While some centres make use of the CHOP INTEND, some experts were of the opinion that this tool is more relevant in a research setting than in Canadian practices and would require additional resources to routinely administer to patients. Similarly, while individual items of the BSID-III are commonly conducted, the full set of the BSID-III and a calculated score is not.

• The definition of hospitalization may provide little value in understanding the health care utilization outside the study

  • Hospitalization is defined in FIREFISH Part 2 as any admission for 2 or more days for any reason that is not due to the study treatment. It is not clear why admissions of a duration less than 2 days were not considered relevant. Also, considering the open-label, single-arm design of FIREFISH Part 2, there is potential for a biased determination of what constitutes a hospitalization that is not due to the study treatment.

Critical Appraisal: SUNFISH Part 2

Internal Validity

SUNFISH Part 2 was a double-blind, placebo-controlled trial that aimed to assess the efficacy of risdiplam in SMA type II patients and non-ambulatory SMA type III patients. Patients were randomized on a 2:1 ratio to risdiplam and placebo, respectively, using an interactive response system. Randomization was stratified by age groups (2 years to 5 years, 6 years to 11 years, 12 years to 17 years, and 18 years to 25 years). The study measured motor function using several tools, with multiple analyses based on outcome measurement data from each tool. The study instituted a gatekeeping hierarchy approach to adjust for multiple testing and to control the type I error rate.

Limitations to the internal validity of SUNFISH Part 2 revolve around the following points:

• Unequal randomization ratio

  • The sponsor randomized patients in a 2:1 ratio to risdiplam or placebo, respectively. Potential challenges that may be associated with such an allocation ratio include the need for larger sample size to capture differences in treatment effect in secondary outcomes, and the potential for reducing the effectiveness of blinding as investigators and assessors would be aware that the probability of being allocated to active treatment is twice that of control. A reduction in statistical power due to the 2:1 randomization ratio in the secondary outcomes cannot be ruled out as an a priori power calculation was only performed for the primary outcome.

• Unequal number of randomized patients in each age group

  • The sponsor’s original plan aimed to randomize a total of 168 patients, stratified by age groups (2 years to 5 years, 6 years to 11 years, 12 years to 17 years, and 18 years to 25 years at randomization). However, only a maximum of 30 patients were to be included and randomized in the age group of 18 years to 25 years, while a minimum of 45 patients were to be randomized into each of the other age groups. The study ended up randomizing 180 patients, of which only 22 were in the age group of 18 years to 25 years. This discrepancy in the number of patients in the age groups means that the age group of 18 years to 25 years did not contribute the same magnitude of effect to the assessed outcomes as the other age groups. In addition, the clinical experts consulted on this review identified this discrepancy as a source of bias toward the younger patients who are biologically more likely to benefit from the treatment.

• Imbalance in the baseline characteristics

  • Some of the more notable discrepancies in baseline characteristics between the risdiplam arm and the placebo arm was the proportion of patients with 3 SMN2 copies (89.2% in the risdiplam arm and 83.3% in the placebo arm), and in the |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||. However, the clinical experts consulted on this review were of the opinion that these discrepancies are unlikely to have a confounding effect on the results. As such, it is not clear if these discrepancies have a directional bias effect or have any effect at all.

• Handling and assessing the impact of missing data

  • The sponsor provided a clear description of handling missing data for the MFM-32, RULM, and HFMSE outcomes. However, the sponsor only planned to conduct a sensitivity analysis with a different approach to data imputation for the primary outcome. This sensitivity analysis was not reported as the percentage of missing data were below the threshold required in the sponsor’s protocol. There was no clear description on the method of handling missing data in outcomes beyond the MFM-32, RULM, and HFMSE. The SMAIS-related outcomes had a considerable amount of missing data and there was no sensitivity analysis to inform on the impact of the missing data in the SMAIS-related outcomes.

External Validity

The inclusion criteria in SUNFISH Part 2 captured a wide range of SMA patients — specifically, patients diagnosed as type II SMA or non-ambulatory type III SMA with a wide age range from 2 years up until 25 years. The outcomes included several potentially relevant measures, including the proportion of motor function responders and respiratory-related outcomes.

Limitations to the external validity of SUNFISH Part 2 revolve around the following points:

• Limited generalizability to the adult population

  • Only 22 out of the 180 patients randomized were in the age group of 18 years to 25 years, the fewest number among the age groups in the study (11.7% in risdiplam and 13.3% in placebo). This limits the contribution of adult patients to the overall conclusion of the study; it also limits the generalizability from the results to adult patients. This limitation was apparent in the various subgroup analyses, where the point estimate of the treatment effect appeared to be lower than the other age groups with wider CIs spanning the null. There was no clear justification as to why the study was planned to purposefully limit the enrolment of adult patients while bolstering other age groups. Investigating the efficacy of risdiplam in the adult population is of special relevance, considering the unmet need and the irreversible nature of the disease progression. The small number of adult patients and the large uncertainty observed in the subgroup analyses limits the generalizability of the overall result of SUNFISH Part 2 to the adult population.

• Changes in the clinical practice and the lack of direct head-to-head comparison to other disease-modifying agents

  • Although it is understandable that the clinical development program for risdiplam was initiated at a time when disease-modifying agents were not yet available, the current clinical landscape in Canada is one where all SMA patients who meet certain criteria are already on nusinersen, and newly diagnosed patients will have nusinersen as a therapeutic option. The SUNFISH trial cannot inform on the comparative efficacy to nusinersen.

• Lack of the inclusion of ambulatory patients

  • Considering the nature of SMA, where alpha motor neurons are irreversibly lost as disease progresses, patients with higher motor functions may have a greater number of alpha motor neurons than patients who have lost such motor functions. Ambulatory patients may thus exhibit a greater variation in the response than non-ambulatory patients, and the generalizability of the SUNFISH results to this patient population may be limited.

• The use of outcome measures without a well-defined MID and results not meeting proposed MID

  • The tool used for the primary outcome, MFM-32, did not have a well-accepted method of assessing the MID. However, the sponsor has indicated that a change of 3 points or more may translate into either the acquisition of a new function or the improvement in performance of several functions. Based on this, the primary outcome, both in point estimate and upper range of the 95% CI, fails to demonstrate such difference (MFM-32 change from baseline mean difference = 1.55 points [95% CI, 0.30 to 2.81]). Two other key motor function measurement tools that were used by the sponsor had published MIDs: RULM had an MID of 2.9 and HFMSE had an MID of 2. The results of the mean difference versus placebo in both of these outcomes also does not achieve the published MID.

• Strong response in the placebo arm

  • Considering the deteriorating nature of the disease and the irreversible nerve damage to alpha motor neurons, the response observed in the placebo arm would have been expected to be minimal. The presence of a relatively strong response in the placebo arm suggests either the need for validated assessment tools that reflect the disease natural history, a potential learning effect in the used outcomes, and/or further studies into the disease natural history in the population reflected in SUNFISH Part 2. In any case, improvements in motor function scales would need to take into consideration the presence of improvements unrelated to treatment.

• Assessment period of 12 months

  • Considering the lifelong nature of SMA and the narrow SMA population in SUNFISH Part 2, generalizability of the efficacy and safety results beyond the presented 12 months remains uncertain. It is, however, understood that the study is ongoing until it reaches the 24-month mark and will continue into an open-label extension study until the time of drug commercialization.

• The use of motor function tools that are not commonly conducted in clinical practice.

  • The clinical experts consulted on this review indicated that the use of MFM-32 is not common across Canadian practices and may require a change in usual assessment practices to incorporate it.

Indirect Evidence

Objectives and Methods for the Summary of Indirect Evidence

There is an absence of direct head-to-head studies comparing risdiplam against other therapies for patients with SMA; ITCs that include risdiplam can provide information on the comparative effectiveness relative to other therapies. The objective of this section is to summarize and critically appraise the evidence available regarding the indirect comparative efficacy of risdiplam to other treatments of SMA in any available ITC.

Description of Indirect Treatment Comparison

The sponsor-submitted ITC aimed to compare the relative treatment effect of risdiplam versus relevant comparators for the treatment of 2 SMA populations: infantile-onset (type I) SMA with 2 copies of the SNM2 gene, and later-onset SMA. Relevant studies were identified through a systematic search of the literature and ITCs were conducted through an anchored MAIC; where an anchored approach was not possible, an unanchored MAIC was used.

An overview of the included ITC is presented in Table 16.

Methods of the Sponsor-Submitted Indirect Treatment Comparison

Objectives

The objective of the sponsor-submitted ITC was to compare the relative treatment effect of risdiplam versus other disease-modifying treatments for SMA. The lack of head-to-head studies comparing risdiplam to other disease-modifying treatments was provided as a rationale for conducting the ITC. The ITC aimed to provide analyses of outcomes related to motor function, respiratory, HRQoL, and safety. The sponsor-submitted ITC aimed to assess the indirect comparative effectiveness of risdiplam in patients with SMA type I and, in a separate analysis, in patients with other SMA types.

Study Selection Methods

The sponsor-submitted ITC described a systematic literature review process that searched several key bibliographic databases and several supplemental sources. The submitted available material did not provide the search strategy or the date at which the search was conducted. In addition, the process of screening retrieved results and extracting data from retrieved results was not outlined in the submitted material. Assessment of the quality of included randomized controlled trials was done through the use of the National Institute for Health and Care Excellence checklist, while assessment of single-arm trials was done through the Agency for Health care Research and Quality tool for risk of bias assessment for single-arm trials. However, the approach for handling poor-quality studies was not outlined in the submitted material.

Efficacy analysis within the SMA type I population included motor function outcomes as measured by the HINE Section 2 and the CHOP INTEND tools, overall survival, and event-free survival. Efficacy analysis in the patient population with type II SMA or type III SMA included motor function outcomes as measured by the HFMSE and RULM tools. For each outcome of interest, the publication must report Kaplan–Meier curves for time-to-event outcomes, mean and/or median for continuous outcomes, or the proportion of individuals for binary outcomes.

Indirect Treatment Comparison Analysis Methods

The sponsor’s overall approach used the MAIC methodology. MAIC is a method used to adjust for between-trial heterogeneity related to differences in baseline patient characteristics to reduce bias in the estimated relative effects between 2 interventions. Under this approach, individual patient-level data from risdiplam studies were weighted to produce a risdiplam-treated sample that resembles the aggregate baseline characteristics of the study of the comparator drug.

Three different MAIC approaches to produce an indirect comparison were outlined in the sponsor-submitted ITC: unanchored MAIC, anchored MAIC, and restricted Bayesian NMA.

Unanchored Matching-Adjusted Indirect Comparison

Unanchored MAIC was needed for the analyses that included the FIREFISH study. Considering the single-arm nature of FIREFISH, there was no possibility to anchor the comparison through a common arm between the risdiplam studies and the comparator study. As such, an unanchored MAIC was performed.

For this analysis, the sponsor treated patients who received the therapeutic dose of risdiplam in the phase II Part 1 of FIREFISH (N = 21) as if they were enrolled in the same study as patients who were in the confirmatory phase III Part 2 of FIREFISH (N = 41). There was no clear description of a method to address potential heterogeneity due to variation in the study design between the 2 parts.

In cases where the comparator study had a shorter duration than FIREFISH, only assessments in FIREFISH conducted at a certain point before the clinical cut-off date were to be included to match the follow-up period of the comparator study.

To determine the effect modifiers and prognostic factors that needed to be included in the weighting model for the indirect comparison, the sponsor conducted a literature search, coupled with a discussion with clinical experts and the internal team. The sponsor ended up adjusting for the following effect modifiers and prognostic factors: mean age at first dose, duration of symptoms, and mean CHOP INTEND score. These were chosen from a list of identified effect modifiers and prognostic factors. Excluded effect modifiers and prognostic factors included mean age at symptom onset, mean age at diagnosis, mean weight, gender, race, mean HINE Section 2 score, ulnar compound muscle action potential (CMAP) amplitude, and peroneal CMAP amplitude. Decisions for the inclusion of effect modifiers and prognostic factors were determined based on the availability of the effect modifiers and prognostic factors in the studies; other reasons for dropped factors were related to a determination by the sponsor that they were already reflected in 1 or more included factors.

A weight for each patient in the FIREFISH study was constructed through a log weight linear regression model where regression parameters were estimated using the method of moments to match effect modifier distribution between trials. Outcomes of patients in the FIREFISH study were then reweighted according to the weight estimates from the model. CIs were obtained through bootstrapping on the whole MAIC process. The sponsor considered methods for estimating the systematic error in the estimates but was unable to implement these approaches due to the small sample size of trials.

Anchored MAIC

An anchored MAIC was conducted for analyses that included the SUNFISH Part 2 study, considering the existence of the placebo arm. When considering the nusinersen study, CHERISH, an assumption was made that placebo was comparable to sham. Based on this assumption, the sponsor aimed to include relevant effect modifiers in the model, but not prognostic factors, as an anchored approach should, theoretically, address the issue of prognostic factors.

The sponsor-submitted ITC chose a subsample of SUNFISH Part 2 in the analysis in an effort to match the inclusion and exclusion criteria of the comparator study (CHERISH).

To determine the effect modifiers that needed to be included in the weighting model for the indirect comparison, the sponsor conducted a literature search, coupled with a discussion with clinical experts and the internal team. The sponsor ended up adjusting for the following effect modifiers: age at screening, baseline motor function score, and SMN2 copy number. These were chosen from a list of identified effect modifiers. Excluded effect modifiers included age at symptom onset, age at diagnosis, disease duration, severe contractures, gender, race, motor milestones achieved, WHO motor milestones achieved, and geographic location. Decisions for the inclusion of effect modifiers were determined based on the availability of the effect modifiers in the studies; other reasons for dropped factors were related to a determination by the sponsor that they were already reflected in 1 or more included factors.

A weight for each patient in the SUNFISH Part 2 study was constructed through a log weight linear regression model where regression parameters were estimated using the method of moments to match effect modifier distribution between trials. Outcomes of patients in the SUNFISH Part 2 study were then reweighted according to the weight estimates by the model. Matching was only conducted on mean baseline characteristics and not on variability, due to the limited number of patients in SUNFISH Part 2. CIs were obtained through bootstrapping on the whole MAIC process.

Restricted Network Meta-Analysis

This was conducted as an alternative approach to the anchored MAIC. The Bayesian approach was performed through a Markov chain Monte Carlo technique. The analysis was performed under both fixed- and random-effects models, where the lower value of the diagnostic information criterion determined the best model fit. Study and treatment-specific terms were given vague priors, while efficacy end points under the random-effects model were given published informative priors due to a paucity of data. The analysis was initiated on 3 chains with different initial values and convergence was assessed using the Brooks-Gelman-Rubin plots. The number of burn-ins, the number of iterations, the assessment of residual deviance, and the assessment of inconsistency were not outlined in the submitted material.

Results of Sponsor-Submitted Indirect Treatment Comparison

Summary of Included Studies

The literature search produced a total of 772 records, of which 64 studies underwent full-text screening. Ultimately, 5 studies were included. Of these 5 studies, 3 addressed the population of infantile-onset type I SMA: FIREFISH (risdiplam), ENDEAR (nusinersen), and STR1VE-US (onasemnogene abeparvovec). The remaining 2 studies addressed SMA type II and SMA type III (symptom onset beyond the infantile period): SUNFISH (risdiplam) and CHERISH (nusinersen).

SMA Type I–Related Studies

The ITC of risdiplam to nusinersen was informed by the FIREFISH and ENDEAR studies. The 2 studies had different designs: FIREFISH consisted of a subgroup of patients who received the high dose in the dose-finding Part 1 pooled with patients from the single-arm design of the confirmatory Part 2 study, while ENDEAR was a double-blind, sham-controlled trial. FIREFISH is an ongoing trial with data available for month 12, and ENDEAR was terminated early and no results were reported at 12 months of treatment, with a median follow-up time in the nusinersen arm of 280 days. In general, enrolment criteria were similar between the 2 trials. Both had similar age restrictions and genetic confirmation of disease, and required infants to have 2 copies of the SMN2 gene.

Overall, the patient characteristics of risdiplam were similar to those of the ENDEAR trial in terms of age and disease duration. The severity of disease at baseline was worse in the FIREFISH patients (CHOP INTEND mean = 22.47 [SD = 6.79] points) when compared with the nusinersen-treated patients (CHOP INTEND mean = 26.63 [SD = 8.13] points), with a similar trend observed in the HINE Section 2 score and in the mean CMAP negative peak amplitude.

To address the discrepancy in the follow-up time, a modified dataset of FIREFISH with a median time in the study of 283 days was used in the base-case analyses of binary end points. In this dataset, any events occurring 6 months before the data cut-off were not considered. For analyses of time-to-event end points (event-free survival, overall survival), all available data from the 12-month data cuts were used.

Outcome definition was similar in terms of the definition of motor responder using the HINE Section 2 score as well as the definition of permanent ventilation.

Although the sponsor-submitted ITC included the STR1VE-US study, unanchored MAIC analysis was not feasible due to lack of overlap between patient populations. As such, a description of the characteristics of STR1VE-US contrasted with FIREFISH is not summarized here.

The sponsor-submitted ITC reported the quality of the included studies as “low risk of bias” or “low to unclear risk of bias.”

SMA Type II– and Type III–Related Studies

The ITC of risdiplam to placebo was informed through the SUNFISH Part 2 (N = 180) and CHERISH (N = 126) studies. Both studies were double-blind, randomized controlled trials. SUNFISH Part 2 was a placebo-controlled study as opposed to the sham-controlled CHERISH study. In addition, SUNFISH is an ongoing study with results available for month 12 of treatment, while CHERISH is a completed study with results of a total follow-up period of 15 months.

There were considerable differences in the inclusion and exclusion criteria of these 2 studies. Specifically, SUNFISH allowed patients up to 25 years of age to enrol, while in CHERISH, only ages up to 12 years were eligible. Restrictions on HFMSE score (a score between 10 and 54) were applied to CHERISH, while no such restrictions were made with SUNFISH. CHERISH excluded patients with severe scoliosis or severe contractures and patients requiring non-invasive ventilation for more than 6 hours during a 24-hour period; however, these exclusion criteria were not applied in SUNFISH.

Baseline characteristics are reflective of the differences in the inclusion and exclusion criteria. Patients in SUNFISH were much older at screening, had a considerably longer disease duration, and a lower HFMSE score. The mean age at symptom onset was similar; however, the range of ages was larger in SUNFISH compared to CHERISH. Gender, RULM score, and SMN2 copy number were similar.

To address the clinical heterogeneity between the 2 trials, only a subset of SUNFISH Part 2 patients was selected according to the following criteria:

• age at screening of 9 years or younger (the maximum age enrolled in CHERISH was 9 years)

• an HFMSE baseline score of 10 or more (an HFMSE score of 10 or more was a CHERISH inclusion criterion)

• no severe scoliosis (the presence of severe scoliosis was a CHERISH exclusion criterion).

In total, 38% of patients enrolled in SUNFISH Part 2 met the subset criteria. Compared to CHERISH, the subset patients were older, had a longer disease duration, and had higher HFMSE and RULM scores. The sponsor-submitted ITC indicated that there were other differences for which there was no way to restrict the subset. These include the presence of severe contractures, difference methods to define independent sitting, and the exclusion of patients requiring non-invasive ventilation of more than 6 hours in CHERISH.

The definition of end points was similar between the studies as both used the RULM and HFMSE scores. The sponsor-submitted ITC report extracted CHERISH outcomes at month 12 where available.

The sponsor-submitted ITC reported the quality of the included studies as “low risk of bias” or “low to unclear risk of bias.”

Results

SMA Type I

The evidence network of the studies informing on the indirect comparison of risdiplam to other disease-modifying therapies is presented in Figure 9. No statistical analysis was possible for the indirect comparison of risdiplam with onasemnogene abeparvovec due to the small overlapping sample post-adjustment.

Subsequent to matching based on the mean age at first dose, mean disease duration at screening, and mean score on CHOP INTEND, an effective sample size of 36.5 was estimated out of the original 58 patients pooled into FIREFISH. The adjusted population had a higher proportion of female patients, a lower proportion of patients with ventilatory support, and a lower mean HINE Section 2 score than patients in the ENDEAR study. These changes are outlined in Table 17.

Ventilation-free survival indicated a hazard ratio of 0.20 (95% CI, 0.06 to 0.42) of risdiplam versus nusinersen. Similarly, overall survival indicated a hazard ratio of 0.26 (95% CI, 0.03 to 0.66). Motor function assessment using the HINE Section 2 showed favourable results for risdiplam in the outcomes of motor milestone response, full head control, and sitting without support, while the outcome of rolling was favourable in the direction of nusinersen. Two outcomes, sitting with and without support and standing, did not show a clear direction. However, all the HINE Section 2–related outcomes had wide CIs, indicating poor statistical robustness in the data. The results are summarized in Table 18.

Beyond weight histograms, no assessment of model diagnostics or residual bias were provided.

Figure 9: Network of Studies Informing Indirect Treatment Comparison in Patients With Type I SMA

SMA = spinal muscular atrophy.

Source: Drug Reimbursement Review sponsor submission.¹⁴

SMA Type II and SMA Type III

The evidence network of the studies informing on the indirect comparison of risdiplam to other disease-modifying therapies is presented in Figure 10.

Subsequent to anchored matching based on the mean age at screening, mean RULM at baseline, and mean SMN2 copy number, an effective sample size of 28.3 in the risdiplam arm and 8.8 in the placebo arm was estimated. Baseline characteristics before and after weighting are presented in Table 19.

Results of the anchored MAIC analysis produced unrealistically wide CIs, indicating a lack of statistical robustness to provide a valid result. Sensitivity analyses using the restricted NMA approach as well as the Bucher method did not provide a clear result that would favour 1 intervention over another and were accompanied with a wide CI. A summary of the obtained MAIC and NMA results is presented in Table 19.

Beyond weight histograms, no assessment of model diagnostics or residual bias were provided.

Figure 10: Network of Studies Informing Indirect Treatment Comparison in Patients With SMA Type II or SMA Type III

SMA = spinal muscular atrophy.

Source: Drug Reimbursement Review sponsor submission.¹⁴

Critical Appraisal of Sponsor-Submitted Indirect Treatment Comparison

The sponsor-submitted ITC described approaches for identifying and analyzing evidence. It also aimed to minimize clinical heterogeneity by analyzing the infantile-onset type I SMA in a separate network and by using an MAIC approach. There were aspects related to the conduct of the systematic review that were not reported. These included literature search, screening method, data extraction method, and handling of poor-quality studies.

The sponsor-submitted ITC identified 3 studies in relation to type I SMA and 2 studies in relation to type II SMA or type III SMA. The network of evidence for type I SMA lacked a common comparator. As such, an unanchored MAIC approach was used to estimate the indirect comparison. For the type II SMA or type III SMA, the sponsor-submitted ITC assumed that the sham procedure is similar to placebo and used this as a common comparator to conduct an anchored MAIC with a sensitivity analysis using a Bayesian NMA approach. And while not described in the methods section, the sponsor also conducted a Bucher-adjusted indirect comparison as another form of sensitivity analysis for the type II SMA and type III SMA network.

To identify effect modifiers and prognostic factors, the sponsor-submitted ITC reported conducting a literature search and a consultation process with clinical experts. In addition, the sponsor-submitted ITC provided a list of all identified factors and the justification to either include or exclude each factor from the MAIC model.

Limitations related to the sponsor-submitted ITC included the following points:

• The use of unanchored MAIC without assessment of residual bias

  • An unanchored MAIC depends on the strong assumption that all known and unknown effect modifiers and prognostic factors are accounted for within the model. This is a very strong assumption that we know has not been achieved. There were several factors that the sponsor-submitted ITC identified as important but did not include. In addition, the differences in the study design between FIREFISH and ENDEAR is another factor that is not accounted for. As such, there is considerable uncertainty regarding the actual observed difference in treatment effect that is attributed to risdiplam. And since the sponsor was unable to assess residual bias due to small sample size, there is no way to estimate the magnitude of the bias due to effect modifiers and prognostic factors in the reported estimates. This limitation is relevant to all outcomes that were estimated from the SMA type I evidence network.

• Lack of clinically homogeneous population post-adjustment

  • The adjusted population had a higher proportion of female patients, a lower proportion of patients with ventilatory support, and a lower mean HINE Section 2 score than patients in the ENDEAR study. These persistent differences may indicate that the resulting population is still not sufficiently similar as to allow valid indirect comparison.

• Improper pooling of FIREFISH study

  • To increase the sample size used in the analysis, the sponsor-submitted ITC pooled patients from FIREFISH Part 1 who received the high (pivotal) dose into the FIREFISH Part 2 patients. FIREFISH Part 1 and FIREFISH Part 2 are 2 independent studies that are not identical in design or in aim. As such, the effect of potential variations in study design, outcome definitions, follow-up times, and patient characteristics is not addressed.

• Small sample size to robustly capture motor function–related differences between risdiplam and nusinersen

  • Examining the results of the ITC in the HINE Section 2 outcomes comparing risdiplam with nusinersen reveals wide CIs that may indicate low statistical robustness of the data or the model. The wide CI combined with the lack of assessment of residual bias further increases the uncertainty associated with the HINE Section 2 results.

• Insufficient sample size to estimate the indirect results of risdiplam versus onasemnogene abeparvovec

  • No sufficient overlapping patients from FIREFISH to STR1VE-US were available to conduct an analysis. Thus, no ITC was estimated for nusinersen versus onasemnogene abeparvovec.

• Considerable differences in the inclusion and exclusion criteria between CHERISH and SUNFISH

  • These differences included age range, restrictions on ventilatory support, restrictions on baseline motor function scores, and restrictions over the existence of anatomical complications due to SMA. The sponsor-submitted ITC attempted to account for these differences through choosing a subset of patients for the SUNFISH study that would match the inclusion and exclusion criteria of CHERISH, but the resulting subset exhibited baseline characteristics that were not as similar to CHERISH as would have been expected. The sponsor-submitted ITC outlined that several other discrepancies in the eligibility criteria remained that were not possible to apply to the SUNFISH Part 2 patients. This limitation is relevant to all outcomes that were estimated from the SMA type II or SMA type III network.

  • In addition, there remained several clinically heterogeneous differences subsequent to adjusting. These included the presence of severe contractures, difference methods to define independent sitting, and the exclusion of patients requiring non-invasive ventilation of more than 6 hours in CHERISH.

• The use of a subset of patients from the SUNFISH study

  • SUNFISH was a randomized, placebo-controlled trial. The main limitation of artificially selecting a subset that is further adjusted for is the potential for increased risk of imbalances within SUNFISH in observed or unobserved factors. This is relevant to the indirect treatment estimates since the placebo group is serving as an anchor in the MAIC, or a bridging node in the NMA analysis approach. This limitation is relevant to all outcomes that were estimated from the SMA type II or SMA type III network.

• Small effective sample size resulting in unrealistically wide CIs from the ITCs in the SMA type II or SMA type III network

  • The wide CIs indicate a lack of statistical precision and prevent any useful interpretation of the presented estimates. The resulting effective sample size was 28.3. This limitation is relevant to all outcomes that were estimated from the SMA type II or SMA type III network.

Summary

One sponsor-submitted ITC was reviewed. The sponsor-submitted ITC compared risdiplam to nusinersen in 2 distinct patient populations: infantile-onset SMA (SMA type I) and later-onset SMA (SMA type II or SMA type III).

Comparison in the SMA type I population was done through an unanchored MAIC between a pooled FIREFISH for and ENDEAR studies. Study design between these 2 studies was different, inclusion and exclusion criteria were similar, and patients in the FIREFISH study were weighted to match the mean age at first dose, duration of symptoms, and the mean CHOP INTEND baseline score of the ENDEAR study. The results of this unanchored MAIC suggest a hazard ratio of ventilation-free survival of risdiplam versus nusinersen of 0.20 (95% CI, 0.06 to 0.42), and an overall survival hazard ratio of 0.26 (95% CI, 0.03 to 0.66). Since this result is the product of an unanchored MAIC with limited adjustment for effect modifiers and prognostic factors, imbalances in trial arms after adjustment, with no assessment of residual bias, it is not clear to what extent is the magnitude of effect due to true treatment difference or to potential measured or unmeasured effect modifiers and prognostic factors. The clinical experts consulted on this review have suggested that there may be a biological rationale to hypothesize why risdiplam could be superior to nusinersen. Specifically, the clinical experts argued that the achievement of drug concentration in the body at a therapeutic level may be faster with risdiplam than nusinersen.

Comparison in the SMA type II or SMA type III population was done through an anchored MAIC along with a sensitivity analysis using a Bayesian NMA approach. Due to discrepancies in the inclusion and exclusion criteria, the sponsor-submitted ITC only used a subset of patients from the SUNFISH Part 2 study that would have been included in the CHERISH study, reducing the sample size of SUNFISH Part 2 by 62%. Subsequently, the weighting of the SUNFISH Part 2 sample produced an effective sample size in the risdiplam arm of 28.3. This small sample size translated into results with wide, and sometimes unrealistic, CIs in both the base-anchored base analysis and the NMA sensitivity analysis. This lack of statistical robustness, as well as the limitations associated with the persistent heterogeneity and the loss of the balanced distribution of characters in the SUNFISH Part 2 after the subset selection, prevents any use of the presented outcomes in a decision-making process.

Other Relevant Evidence

This section includes 4 submitted phase II studies and additional relevant studies included in the sponsor’s submission to CADTH that were considered to address important gaps in the evidence included in the systematic review.

FIREFISH Part 1 provides additional information on the safety and efficacy of risdiplam in infants with SMA type I. SUNFISH Part 1 provides additional information on the safety and efficacy of risdiplam in infant and young adult patients with SMA type II or SMA type III. JEWELFISH will provide evidence of the safety of risdiplam in pediatric and young adult patients with SMA who have been previously treated with RO6885247, nusinersen, olesoxime, or onasemnogene abeparvovec. RAINBOWFISH will provide additional evidence on the safety and efficacy of risdiplam in infants who have been genetically diagnosed with SMA but are pre-symptomatic. The JEWELFISH and RAINBOWFISH studies are ongoing.

Phase II Clinical Trials: FIREFISH Part 1 and SUNFISH Part 1

Methods

FIREFISH Part 1 was a multi-centre, open-label, single-arm, phase II study of the safety, tolerability, PK, and PD of risdiplam for the treatment of infants with SMA type I. It preceded FIREFISH Part 2, which has been described in the main report. Part 1 was conducted at 7 sites in 5 countries (Belgium, France, Italy, Switzerland, and the US). Enrolment began in December 2016 for infants (aged 1 month to 7 months) with SMA type I and included 21 patients separated into 2 cohorts for different dosing targets: 700 ng*hour/mL and 2,000 ng*hour/mL or less. Once doses were selected for FIREFISH Part 2, patients could continue receiving the appropriate dose for a total 24-month treatment period followed by an additional extension period until risdiplam is commercially available or is no longer produced by the sponsor (Figure 11). The selected doses were 0.04 mg/kg for infants from 1 month to 3 months of age, 0.08 mg/kg for infants between 3 months and 5 months, and 0.2 mg/kg for infants older than 5 months.

SUNFISH Part 1 was a multi-centre, open-label, placebo-controlled, dose-ranging, phase II study of the safety, tolerability, PK, and PD of risdiplam in infant and young adult patients with SMA type II or SMA type III. It preceded SUNFISH Part 2, which has been described in the main report. Part 1 was conducted at 5 sites in 4 countries (Italy, Germany, France, and Belgium). Enrolment began in October 2016 for patients (aged 2 years to 25 years) with SMA type II or SMA type III and included 51 patients separated into 2 groups by age: 2 years to 11 years and 12 years to 25 years. During the initial minimum 12-week, double-blind, placebo-controlled phase, patients in the younger cohort could receive 1 of the 0.02 mg/kg, 0.05 mg/kg, or 0.25 mg/kg doses while those in the older cohort could receive either 3 mg or 5 mg doses (Figure 12). After 12 weeks, patients on placebo were switched to active treatment in the dose group they were assigned until pivotal doses were selected. The selected doses were 5 mg if the patient’s body weight was at least 20 kg and 0.25 mg/kg if body weight was less than 20 kg. All patients then received the pivotal dose for a 24-month treatment period followed by an additional extension period until risdiplam is commercially available or is no longer produced by the sponsor. Of the 51 patients, 14 received the pivotal dose from the beginning of treatment, 21 started active treatment at a lower-dose before switching to the pivotal dose, and 16 started on placebo before switching to a lower dose initially (n = 10) or directly receiving the pivotal dose (n = 6).

Figure 11: Study Design of FIREFISH

AUC_0-24,ss = area under the curve from 0 hours to 24 hours at steady state; PK = pharmacokinetics; SMN2 = survival of motor neuron 2.

Source: Clinical Study Report of FIREFISH.⁴³

Figure 12: Study Design of SUNFISH Part 1

DB = double-blind; OLE = open-label extension.

Source: Clinical Study Report of SUNFISH.⁴⁴

Populations

The inclusion and exclusion criteria for FIREFISH Part 1 and SUNFISH Part 1 are outlined in Table 21.

In FIREFISH Part 1, patients were separated into 2 cohorts based on which dose of risdiplam they received. Patients in cohort 1 received dose level 1 (dose of risdiplam that would achieve a target exposure of 700 ng*hour/mL) while those in cohort 2 received dose level 2 (target exposure of ≤ 2,000 ng*hour/mL). Table 22 summarizes the baseline characteristics of patients in FIREFISH Part 1. Briefly, the mean (SD) age was 5.81 (1.38) months at enrolment and 71.4% of patients were female. The majority (81%) were White and the rest were Asian. A proportion of patients was receiving ventilation support (14.3% received BiPAP for fewer than 16 hours per day and 19.0% received ventilation prophylactically) and most (95.2%) were able to swallow.

In SUNFISH Part 1, patients aged 12 years to 25 years made up group A and patients aged 2 years to 11 years made up group B. Table 23 summarizes the baseline characteristics of patients in SUNFISH Part 1. Briefly, the younger cohort had a mean (SD) age of || |||| years while the older cohort had a mean (SD) age of |||||||||| years at screening. Just over half of all patients were female |||||||||||||||||||||| were White. Overall, |||| of patients were receiving fewer than 16 hours of BiPAP support per day (|||||||||| |||| in the younger and older cohorts, respectively) and |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| of the older patients were not. Few patients (13.7%) were ambulatory: 19.4% of those aged 2 years to 11 years and 5% of those aged 12 years to 25 years.